You can deploy this model and start earning money today!

Views

No views yet

Website

📄 Paper | 📝 Blog | 📚 Collection

ColBERT-Zero

Name: ColBERT-Zero-supervised
Rating: 2.5 (87 reviews)

🎯 TL;DR: First large-scale fully pre-trained ColBERT model using only public data. Achieves 55.43 nDCG@10 on BEIR benchmark, outperforming GTE-ModernColBERT and GTE-ModernBERT trained on closed and stronger data. New SOTA on BEIR for models <150M parameters.

Why ColBERT-Zero?

Late interaction (ColBERT / multi-vector) models have clear advantages in out-of-domain generalization, long-context handling, and reasoning-intensive retrieval. Yet they remain undertrained: current state-of-the-art ColBERT models (e.g, GTE-ModernColBERT and ColBERT-small) are simply built by bolting a small knowledge distillation step onto a strong dense (single-vector) model. Even recent efforts like mxbai-edge-colbert-v0 perform all early training stages in a single-vector setting, only switching to the multi-vector objective at the very end.

This leaves a lot of performance on the table. ColBERT-Zero demonstrates that performing contrastive pre-training directly in the multi-vector setting, rather than treating it as an afterthought, unlocks a significantly higher performance ceiling. Trained exclusively on public data (Nomic-embed dataset mixture), ColBERT-Zero overcomes a 2.4-point data quality disadvantage to outperform models trained on proprietary, closed-source data. For detailed results, please have a look at our blogpost and the paper. All the models (including intermediate checkpoints) as well training code are released under an Apache 2.0 license.

Controlled Comparison Design

We deliberately trained on the public Nomic-embed data mixture for a strategic reason: Nomic has already trained a dense ModernBERT model (ModernBERT-embed) on this exact data. This lets us compare dense vs. multi-vector training with the same data, same base model (ModernBERT), and same pipeline. The only variable is whether the contrastive phases are performed in the dense or multi-vector setting.

This design reveals a striking result: the dense baseline trained on Nomic data scores 52.89, while the one trained on GTE's proprietary data scores 55.33: a 2.4-point data quality gap. Despite this disadvantage, ColBERT-Zero's full multi-vector pre-training pipeline closes and surpasses this gap, reaching 55.43 nDCG@10.

The Three-Phase Training Pipeline

The development followed a three-phase pipeline, each providing a different type of learning signal:

Phase 1 - Unsupervised Contrastive Pre-training

We began with the nomic-embed-unsupervised-data dataset. Using PyLate's GradCache implementation to scale per-GPU batch size without VRAM constraints, combined with cross-GPU gathering of representations, we reached effective batch sizes of ~16k, required for unsupervised training to produce plausible in-batch hard negatives. Unlike dense training, the multi-vector objective allows the encoder to learn fine-grained token importance from the very first phase.

Phase 2 - Supervised Contrastive Fine-tuning

We refined the model using the nomic-embed-supervised-data. This stage introduced mined hard negatives: documents that are superficially similar to the query but not actually relevant. This allows teaching the model to handle nuance by prioritizing specific keywords and contextual tokens most indicative of a true match.

Phase 3 - Knowledge Distillation (KD)

The final stage used the ms-marco-en-bge dataset. We leveraged a powerful Gemma-based model as a teacher, allowing our student models to learn to replicate complex reasoning scores via the efficient MaxSim operator.

Key Findings

1. The Standard Recipe Leaves Performance on the Table

The KD-only approach (the current industry standard) scores 54.09, lagging behind full pre-training by 1.3 points. A simple distillation step is insufficient for optimal multi-vector performance.

2. Supervised + KD Is the Efficiency Sweet Spot

By running a supervised contrastive step in the multi-vector setting before distillation, we reach 55.12 nDCG@10, closing most of the gap with the fully pre-trained model (55.43). This costs ~40 GH200-hours instead of ~408: roughly 10× cheaper for 99.4% of the performance.

3. Prompt Alignment Is Non-Negotiable

Nomic's base models are pre-trained with asymmetric prompts (search_query: and search_document:). While ColBERT has its own asymmetric mechanism via [Q] and [D] markers, we found:

Stripping pre-training prompts during fine-tuning causes significant performance degradation.
Adding prompts to a model not pre-trained with them also hurts performance.
Even with perfect alignment, prompts provide an intrinsic benefit: full ColBERT pre-training with prompts (55.43) vs. without prompts (54.61), no mismatch in either case, shows a meaningful 0.82-point gap.

Why do prompts help? Our leading hypothesis is that prompt tokens act as implicit query expansion: extra slots that don't carry specific meaning but let the model store global information about the sequence. The original ColBERT used [PAD] tokens for this purpose, but modern Flash Attention implementations broke this trick (masked tokens no longer produce usable embeddings). Explicit prompt tokens may be quietly re-enabling it.

Practical takeaway: Always align your prompts with the base model's pre-training setup. Misalignment is one of the easiest ways to silently lose performance. Note that this sensitivity decreases with stronger downstream fine-tuning: with enough training, the model can adapt to an initial mismatch.

Model Lineup

The Main Models (ColBERT-Zero)

ColBERT-Zero utilizes the full 3-phase pipeline with strict prompt alignment, achieving 55.43 nDCG@10 on BEIR, setting a new SOTA for models <150M parameters. We also provide ColBERT-Zero-noprompts, the same pipeline without asymmetric prompts, to study the impact of query expansion on multi-vector performance.

The cheap-to-train ones (ModernColBERT-embed-base)

These models represent the practical sweet spot. By skipping the expensive unsupervised phase, ModernColBERT-embed-base (Supervised + KD) achieves ~97% of the flagship's performance at only ~10% of the compute cost. For reference, ModernColBERT-embed-base-kd performs only the distillation step on a supervised dense base.

Intermediate Checkpoints

For researchers studying the incremental impact of each phase and prompt alignment, we release several ablation variants: ColBERT-Zero-supervised, ColBERT-Zero-unsupervised (and their -noprompts versions), and ModernColBERT-embed-base-supervised.

Full Performance on BEIR

Model	Avg	FiQA	NFCorpus	TREC-COVID	Touche	ArguAna	Quora	SCIDOCS	SciFact	NQ	ClimateFEVER	HotpotQA	DBPedia	CQADupstack	FEVER	MSMARCO
Baselines
ModernBERT-embed-unsupervised	47.05	42.53	35.33	68.44	18.58	48.82	88.63	19.83	72.30	46.32	22.97	60.00	37.97	42.40	67.39	34.23
ModernBERT-embed-supervised	52.89	40.59	33.40	84.15	31.91	48.96	88.85	18.59	69.63	62.15	35.67	67.11	41.50	42.08	87.35	41.47
GTE-ModernColBERT	54.67	45.28	37.93	83.59	31.23	48.51	86.61	19.06	76.34	61.80	30.62	77.32	48.03	41.00	87.44	45.32
gte-modernbert-base	55.33	48.81	36.44	81.95	21.68	72.68	88.55	21.29	77.40	57.62	37.74	69.47	41.79	42.63	91.03	40.90
KD from dense supervised
ModernColBERT-embed-base-kd-only	54.09	42.51	37.01	79.52	34.58	51.75	87.67	18.15	75.04	61.45	28.31	76.70	47.54	40.68	84.82	45.57
Supervised + KD from dense unsupervised
ModernColBERT-embed-base-supervised	50.72	40.09	35.56	71.12	25.53	44.27	86.96	18.19	73.78	58.89	32.95	71.49	43.23	42.55	70.51	45.72
ModernColBERT-embed-base	55.12	41.50	36.51	77.46	33.77	52.45	86.26	18.66	74.90	62.24	37.27	80.07	48.27	41.60	89.71	46.17
ColBERT-Zero
Unsupervised	51.44	45.38	36.88	67.82	22.59	51.53	87.78	22.30	76.76	58.80	24.24	68.29	43.16	45.76	81.58	38.78
Supervised	51.81	42.45	35.60	74.72	23.83	41.81	87.19	19.85	73.71	61.95	35.01	71.37	46.20	45.16	72.61	45.68
Distilled	55.43	42.62	37.28	78.69	36.13	53.07	85.24	19.88	76.50	61.66	35.72	79.41	47.48	41.34	90.59	45.80
ColBERT-Zero-noprompts
Unsupervised	51.70	45.31	34.72	73.55	23.26	52.56	88.15	22.63	76.10	59.18	24.24	66.66	42.61	45.56	81.88	39.15
Supervised	52.39	43.36	36.01	72.42	23.79	47.42	87.79	21.30	73.85	62.25	31.61	70.32	44.07	44.03	85.54	42.11
Distilled	54.61	43.14	36.60	78.60	36.36	49.49	88.05	19.13	76.42	61.73	32.70	76.99	47.69	40.21	85.97	46.01

Limitations & Discussion

Data-specific findings. We deliberately used the Nomic Embed data mixture for controlled comparison. Some observations (particularly around prompt sensitivity) may not generalize to different or stronger training configurations.
Scale vs. objective. The gains from multi-vector pre-training likely reflect more training time in the multi-vector setting, rather than the contrastive objective itself. Performing KD alone at a larger scale might yield similar or superior results due to the higher quality of the distillation signal. Our study uses the conventional setup where training scale is inversely proportional to signal quality, reflecting the higher cost of generating high-quality labels.
Prompt sensitivity decreases with stronger fine-tuning. When experimenting with stronger fine-tuning data (e.g., NV-Retriever), adding prompts on top of a model pre-trained without them did not degrade results the way it did with ColBERT-Zero. With enough downstream training, the model can adapt to an initial mismatch.

Serving at Scale

For production deployment of ColBERT-Zero and other multi-vector models, check out NextPlaid and FastPlaid, our production-grade engines for multi-vector retrieval.

Resources

📦 All checkpoints: HF Collection - every phase, with and without prompts
💻 Code: Training boilerplates
📄 Paper: ArXiv

Model Details

Model Description

Model Type: PyLate model

Document Length: 519 tokens
Query Length: 39 tokens
Output Dimensionality: 128 tokens
Similarity Function: MaxSim
Training Datasets:
- medi_sts_wiki_rephrasal
- fever_hn_mine
- msmarco_distillation_simlm_rescored_reranked_min15
- reddit_triples
- medi_sts_flickr_sampled
- nq_cocondensor_hn_mine_reranked_min15
- nli_simcse_50negs_fixed
- medi_supernli_sampled
- hotpotqa_hn_mine_shuffled
- medi_sts_stackexchange_dupe

Model Sources

Documentation: PyLate Documentation
Repository: PyLate on GitHub
Hugging Face: PyLate models on Hugging Face

Full Model Architecture

ColBERT(
  (0): Transformer({'max_seq_length': 518, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Dense({'in_features': 768, 'out_features': 128, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': False})
)

Usage

First install the PyLate library:

pip install -U pylate

[!WARNING] Prompt alignment is critical for ColBERT-Zero models. You must use prompt_name="query" when encoding queries and prompt_name="document" when encoding documents. ColBERT-Zero was pre-trained with asymmetric prompts (search_query: / search_document:), and stripping them causes significant performance.

Retrieval

Use this model with PyLate to index and retrieve documents. The index uses FastPLAID for efficient similarity search.

Indexing documents

Load the ColBERT model and initialize the PLAID index, then encode and index your documents:

python

from pylate import indexes, models, retrieve

# Step 1: Load the ColBERT model
model = models.ColBERT(
    model_name_or_path="pylate_model_id",
)

# Step 2: Initialize the PLAID index
index = indexes.PLAID(
    index_folder="pylate-index",
    index_name="index",
    override=True,  # This overwrites the existing index if any
)

# Step 3: Encode the documents
documents_ids = ["1", "2", "3"]
documents = ["document 1 text", "document 2 text", "document 3 text"]

documents_embeddings = model.encode(
    documents,
    batch_size=32,
    is_query=False,  # Ensure that it is set to False to indicate that these are documents, not queries
    prompt_name="document", # ⚠️ Required for ColBERT-Zero! Do not omit.
    show_progress_bar=True,
)

# Step 4: Add document embeddings to the index by providing embeddings and corresponding ids
index.add_documents(
    documents_ids=documents_ids,
    documents_embeddings=documents_embeddings,
)

Note that you do not have to recreate the index and encode the documents every time. Once you have created an index and added the documents, you can re-use the index later by loading it:

python

# To load an index, simply instantiate it with the correct folder/name and without overriding it
index = indexes.PLAID(
    index_folder="pylate-index",
    index_name="index",
)

Retrieving top-k documents for queries

Once the documents are indexed, you can retrieve the top-k most relevant documents for a given set of queries. To do so, initialize the ColBERT retriever with the index you want to search in, encode the queries and then retrieve the top-k documents to get the top matches ids and relevance scores:

[!WARNING] Always pass prompt_name="query" for queries and prompt_name="document" for documents. Omitting these prompts will silently degrade retrieval quality.

python

# Step 1: Initialize the ColBERT retriever
retriever = retrieve.ColBERT(index=index)

# Step 2: Encode the queries
queries_embeddings = model.encode(
    ["query for document 3", "query for document 1"],
    batch_size=32,
    is_query=True,  #  # Ensure that it is set to False to indicate that these are queries
    prompt_name="query", # ⚠️ Required for ColBERT-Zero! Do not omit.
    show_progress_bar=True,
)

# Step 3: Retrieve top-k documents
scores = retriever.retrieve(
    queries_embeddings=queries_embeddings,
    k=10,  # Retrieve the top 10 matches for each query
)

Reranking

[!WARNING] Always pass prompt_name="query" for queries and prompt_name="document" for documents. Omitting these prompts will silently degrade retrieval quality.

If you only want to use the ColBERT model to perform reranking on top of your first-stage retrieval pipeline without building an index, you can simply use rank function and pass the queries and documents to rerank:

python

from pylate import rank, models

queries = [
    "query A",
    "query B",
]

documents = [
    ["document A", "document B"],
    ["document 1", "document C", "document B"],
]

documents_ids = [
    [1, 2],
    [1, 3, 2],
]

model = models.ColBERT(
    model_name_or_path="pylate_model_id",
)

queries_embeddings = model.encode(
    queries,
    is_query=True,
    prompt_name="query" # ⚠️ Required for ColBERT-Zero! Do not omit.
)

documents_embeddings = model.encode(
    documents,
    is_query=False,
    prompt_name="document" # ⚠️ Required for ColBERT-Zero! Do not omit.
)

reranked_documents = rank.rerank(
    documents_ids=documents_ids,
    queries_embeddings=queries_embeddings,
    documents_embeddings=documents_embeddings,
)

Evaluation

Metrics

Py Late Information Retrieval

Dataset: ['NanoClimateFEVER', 'NanoDBPedia', 'NanoFEVER', 'NanoFiQA2018', 'NanoHotpotQA', 'NanoMSMARCO', 'NanoNFCorpus', 'NanoNQ', 'NanoQuoraRetrieval', 'NanoSCIDOCS', 'NanoArguAna', 'NanoSciFact', 'NanoTouche2020']
Evaluated with pylate.evaluation.pylate_information_retrieval_evaluator.PyLateInformationRetrievalEvaluator

Metric	NanoClimateFEVER	NanoDBPedia	NanoFEVER	NanoFiQA2018	NanoHotpotQA	NanoMSMARCO	NanoNFCorpus	NanoNQ	NanoQuoraRetrieval	NanoSCIDOCS	NanoArguAna	NanoSciFact	NanoTouche2020
MaxSim_accuracy@1	0.4	0.86	0.88	0.5	0.96	0.58	0.46	0.6	0.88	0.54	0.22	0.7	0.6939
MaxSim_accuracy@3	0.72	0.92	0.96	0.66	1.0	0.76	0.62	0.78	0.98	0.76	0.52	0.82	0.9388
MaxSim_accuracy@5	0.78	0.94	0.98	0.72	1.0	0.84	0.7	0.84	0.98	0.82	0.62	0.82	0.9388
MaxSim_accuracy@10	0.82	0.98	1.0	0.84	1.0	0.9	0.72	0.9	1.0	0.9	0.84	0.86	0.9796
MaxSim_precision@1	0.4	0.86	0.88	0.5	0.96	0.58	0.46	0.6	0.88	0.54	0.22	0.7	0.6939
MaxSim_precision@3	0.28	0.7467	0.3267	0.3267	0.5933	0.2533	0.4133	0.26	0.4	0.42	0.1733	0.3	0.6463
MaxSim_precision@5	0.228	0.676	0.204	0.264	0.368	0.168	0.384	0.172	0.248	0.328	0.124	0.18	0.6041
MaxSim_precision@10	0.136	0.548	0.108	0.156	0.19	0.09	0.296	0.098	0.132	0.22	0.084	0.096	0.5
MaxSim_recall@1	0.21	0.0885	0.8367	0.2826	0.48	0.58	0.0443	0.57	0.7707	0.1147	0.22	0.675	0.05
MaxSim_recall@3	0.354	0.2035	0.9167	0.46	0.89	0.76	0.1	0.73	0.9387	0.2597	0.52	0.805	0.1354
MaxSim_recall@5	0.4523	0.2808	0.9433	0.557	0.92	0.84	0.128	0.79	0.9493	0.3367	0.62	0.805	0.2072
MaxSim_recall@10	0.5207	0.3875	0.9733	0.6865	0.95	0.9	0.1576	0.88	0.986	0.4507	0.84	0.85	0.3284
MaxSim_ndcg@10	0.4515	0.6936	0.9221	0.5661	0.9269	0.7348	0.3706	0.7296	0.9333	0.4419	0.51	0.782	0.5667
MaxSim_mrr@10	0.5562	0.8996	0.9283	0.6025	0.98	0.6822	0.5502	0.6999	0.9333	0.658	0.4071	0.7645	0.8079
MaxSim_map@100	0.3643	0.5441	0.8936	0.4952	0.8907	0.6885	0.1699	0.6749	0.9071	0.3449	0.4144	0.7626	0.4245

Nano BEIR

Dataset: NanoBEIR_mean
Evaluated with pylate.evaluation.nano_beir_evaluator.NanoBEIREvaluator

Metric	Value
MaxSim_accuracy@1	0.6365
MaxSim_accuracy@3	0.803
MaxSim_accuracy@5	0.8445
MaxSim_accuracy@10	0.903
MaxSim_precision@1	0.6365
MaxSim_precision@3	0.3954
MaxSim_precision@5	0.3037
MaxSim_precision@10	0.2042
MaxSim_recall@1	0.3786
MaxSim_recall@3	0.5441
MaxSim_recall@5	0.6023
MaxSim_recall@10	0.6854
MaxSim_ndcg@10	0.6638
MaxSim_mrr@10	0.7284
MaxSim_map@100	0.5827

Training Details

Training Datasets

medi_sts_wiki_rephrasal

Dataset: medi_sts_wiki_rephrasal
Size: 25,000 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 8 tokens mean: 14.19 tokens max: 39 tokens	min: 7 tokens mean: 14.05 tokens max: 39 tokens	min: 7 tokens mean: 14.16 tokens max: 39 tokens	min: 7 tokens mean: 14.08 tokens max: 33 tokens	min: 7 tokens mean: 14.1 tokens max: 39 tokens	min: 7 tokens mean: 14.23 tokens max: 39 tokens	min: 7 tokens mean: 14.19 tokens max: 39 tokens	min: 7 tokens mean: 14.45 tokens max: 39 tokens	min: 7 tokens mean: 14.05 tokens max: 34 tokens	min: 8 tokens mean: 14.1 tokens max: 33 tokens	min: 7 tokens mean: 14.13 tokens max: 39 tokens	min: 7 tokens mean: 14.28 tokens max: 39 tokens	min: 8 tokens mean: 14.21 tokens max: 38 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`What happens in the dehidration process?`	`How does process of exfoliation happen?`	`The main Objective of the phillipine electrical code?`	`What engine will fit to a 87 prelude engine bay?`	`What is the emotion the reader feels from reading a story?`	`How old are you when you read the torah?`	`Melbourne airport to melbourne city centre?`	`What is rahu desa?`	`In mexico what is the wildlife like?`	`This climate is typical of the southern california?`	`Where do you put the product key of microsoft in a computer?`	`What is the product of the wavelength and the frequency?`	`What war had the battle of manila bay?`
`What two rivers in Washington State start with Q?`	`What 2 rivers empty into the Pacific Ocean that start with q?`	`Where is the Bobcat in the food web?`	`What animals live in washington's rainforest?`	`Where is the power steering fluid reservoir on a 325i?`	`A whole number greater than one who's only factors are one and itself?`	`Drive time to santa fe new mexico from denver?`	`How does the US keep illegal immigrants out of the US?`	`How did use somebody by the kings of leon come about?`	`What are themes in the book Stargirl?`	`What is the visual examination of the vagina?`	`What is the voice in Neil Armstrong made history by being the first man on the moon?`	`What chemical change cause pollution how can these changes harm the environment?`
`Why do the halogens in group 7A have such electron affinities?`	`Why do the halogens in group 7a have such high electron in affinities?`	`What are some interesting attractions in georgia?`	`Can human be an immortal vampire?`	`Benjamin franklin created an invention called the lightning rod to help keep houses safe from lightning?`	`Why is'nt it correct?`	`How does science relate in your everyday lives?`	`How many episode in bones season 4?`	`What are the top ten movies ever watched?`	`What does te gusta la clase de matematicas means?`	`What is the lake in the boniva commercial?`	`What do New zealanders call santa claus?`	`What causes mouth twitching?`

Loss: pylate.losses.contrastive.Contrastive

fever_hn_mine

Dataset: fever_hn_mine
Size: 140,085 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 7 tokens mean: 14.39 tokens max: 39 tokens	min: 30 tokens mean: 38.97 tokens max: 39 tokens	min: 6 tokens mean: 37.47 tokens max: 39 tokens	min: 6 tokens mean: 37.49 tokens max: 39 tokens	min: 5 tokens mean: 37.38 tokens max: 39 tokens	min: 4 tokens mean: 37.67 tokens max: 39 tokens	min: 6 tokens mean: 37.86 tokens max: 39 tokens	min: 6 tokens mean: 37.75 tokens max: 39 tokens	min: 5 tokens mean: 37.45 tokens max: 39 tokens	min: 6 tokens mean: 37.66 tokens max: 39 tokens	min: 8 tokens mean: 37.7 tokens max: 39 tokens	min: 6 tokens mean: 37.63 tokens max: 39 tokens	min: 5 tokens mean: 37.49 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`Nikolaj Coster-Waldau worked with the Fox Broadcasting Company.`	Fox Broadcasting Company The Fox Broadcasting Company ( often shortened to Fox and stylized as FOX ) is an American English language commercial broadcast television network that is owned by the Fox Entertainment Group subsidiary of 21st Century Fox . The network is headquartered at the 20th Century Fox studio lot on Pico Boulevard in the Century City section of Los Angeles , with additional major offices and production facilities at the Fox Television Center in nearby West Los Angeles and the Fox Broadcasting Center in the Yorkville neighborhood of Manhattan , New York City . It is the third largest major television network in the world based on total revenues , assets , and international coverage . Launched on October 9 , 1986 as a competitor to the Big Three television networks ( ABC , NBC and CBS ) , Fox went on to become the most successful attempt at a fourth television network . It was the highest-rated broadcast network in the 18 -- 49 demographic from 2004 to 2012 , and earne...	Nikolaj Coster-Waldau Nikolaj Coster-Waldau ( -LSB- neɡ̊olaɪ̯ kʰʌsd̥ɐ ˈʋald̥ɑʊ̯ -RSB- ; born 27 July 1970 ) is a Danish actor , producer and screenwriter . He graduated from Danish National School of Theatre in Copenhagen in 1993 . Coster-Waldau 's breakthrough performance in Denmark was his role in the film Nightwatch ( 1994 ) . Since then he has appeared in numerous films in his native Scandinavia and Europe in general , including Headhunters ( 2011 ) and A Thousand Times Good Night ( 2013 ) . In the United States , his debut film role was in the war film Black Hawk Down ( 2001 ) , playing Medal of Honor recipient Gary Gordon . He then played Detective John Amsterdam in the short-lived Fox television series New Amsterdam ( 2008 ) , as well as appearing as Frank Pike in the 2009 Fox television film Virtuality , originally intended as a pilot . He became widely known to a broad audience for his current role as Ser Jaime Lannister , in the HBO series Game of Thrones . In 2017 , he bec...	`Nukaaka Coster-Waldau Nukaaka Coster-Waldau ( née Sascha Nukaka Motzfeldt ; born 23 February 1971 ) , Nukâka , is a Greenlandic singer , actress , and a former Miss Greenland . She is married to Danish actor Nikolaj Coster-Waldau and has two daughters with him . The Greenlandic actress of Inuit , German and Norwegian extraction was born in Uummannaq , Greenland . She is the daughter of Vivi and Josef Motzfeldt , a Greenlandic politician .`	`Fox Channel (Germany) Fox ( also known as Fox Channel or FOX Channel Germany ) is a German pay-tv-channel dedicated to television drama series . The channel heavily features American and British drama series from the major broadcast networks and premium cable networks . They are broadcast without interruptions with both German and English audio .`	New Amsterdam (TV series) New Amsterdam is an American television drama which aired for eight episodes in 2008 on Fox . The series starred Nikolaj Coster-Waldau as `` John Amsterdam '' ( real name Johann van der Zee ) , an immortal Dutch man born in 1607 , who has lived in New York City on and off since he was 14 years old , and who is a homicide detective in the present day . The series was nominated for an Emmy for Main Title Design .	`Nick Coster Nick Coster ( born 24 September 1985 in Alphen aan den Rijn ) is a Dutch footballer who currently plays for Hoofdklasse side DOVO . He last played professionally as a left back for RKC Waalwijk in the Dutch Eerste Divisie , whom he had joined in 2014 .`	`Nicolas Coster Nicolas Dwynn Coster ( born December 3 , 1933 ) is a British-born American actor , most known for his work in daytime drama and as a character actor on nighttime television series , such as Buck Rogers in the 25th Century and Star Trek : The Next Generation .`	`Claus Spahn Claus Spahn ( born 15 May 1940 in Bottrop , Germany ) is a former TV programme editor of Germany 's Westdeutscher Rundfunk ( WDR ) broadcasting corporation , a presenter , producer and author . He was awarded the Cross of the Order of Merit of the Federal Republic of Germany in 2001 .`	Jonathan Elias Weiske Jonathan-Elias Weiske ( born in Berlin on 4 February 1996 ) is a German film and television actor and voice actor . Jonathan was born in Berlin son of Oliver Betke and Claudia Weiske , an artistic family , with his father being a film/television/theatre actor who had also appeared in musicals in addition to working also a synchronspeecher ( voice over ) and a photographer . Jonathan 's mother also worked as an actress who had appeared amongst others in Gute Zeiten , schlechte Zeiten in the role Elke Opitz . She had also appeared in a number of theater productions in Berlin ( in Theater am Kurfürstendamm ) , Dresden and in Hamburg . Jonathan started as a child actor at age five . He had his screen debut appearing in the series Herzschlag-Die Retterand two years later in the short film Jesus Christ , directed by U. Preatel . He became well-known through the series Fünf Sterne that ran from 2005 to 2008 on ZDF , where he starting as a 9-year old , he played the c...	Fox (Finland) Fox ( formerly SuomiTV ) is a Finnish entertainment TV channel owned and produced by Fox Networks Group ( FNG ) Nordic . The channel was acquired by FNG Nordic in January 2012 , and then relaunched as the current incarnation on 16 April that year . Fox is freely available in Finland both over-the-air and through cable , as it bases its funding on television advertisements . It is FIC Nordic 's first channel that does not operate on subscriber fees . The channel provides a free catch up service through its online platform FOXPlay	Fox News Fox News ( officially known as the Fox News Channel ) is an American basic cable and satellite news entertainment television channel owned by the Fox Entertainment Group , a subsidiary of 21st Century Fox . The channel broadcasts primarily from studios at 1211 Avenue of the Americas , New York City , New York . The channel was created by Australian-American media mogul Rupert Murdoch , who hired former Republican Party media consultant and CNBC executive Roger Ailes as its founding CEO . It launched on October 7 , 1996 , to 17 million cable subscribers . Fox News grew during the late 1990s and 2000s to become a dominant cable news network in the United States . Rupert Murdoch is the current chairman and acting CEO of Fox News . Fox News Channel has been accused of biased reporting and promoting the Republican Party . Critics have cited the channel as detrimental to the integrity of news overall . Fox News employees have responded that news reporting operates independentl...	Prva Srpska Televizija Prva ( -LSB- pr̩̂ːʋaː -RSB- ) or Prva Srpska Televizija ( Прва српска телевизија , in English : First Serbian Television ) is a Serbian commercial television network with national coverage . Launched on 31 December 2006 at 7pm as Fox televizija , the network has since December 2009 been owned by the Greek media company Antenna Group . From its inception until December 2009 , it was majority owned by the global media conglomerate News Corporation , which attached its Fox brand to the operation , naming Dan Bates as the CEO . During its time under News Corporation , Fox televizija closely co-operated with News Corporation owned networks in neighboring countries such as bTV in Bulgaria and Fox Turkey , sharing production capabilities on specific projects and making use of their experiences . Though quickly establishing itself on the Serbian television market and improving its viewership stake year after year , Fox televizija was a money-losing operation , postin...
`Nikolaj Coster-Waldau worked with the Fox Broadcasting Company.`	Nikolaj Coster-Waldau Nikolaj Coster-Waldau ( -LSB- neɡ̊olaɪ̯ kʰʌsd̥ɐ ˈʋald̥ɑʊ̯ -RSB- ; born 27 July 1970 ) is a Danish actor , producer and screenwriter . He graduated from Danish National School of Theatre in Copenhagen in 1993 . Coster-Waldau 's breakthrough performance in Denmark was his role in the film Nightwatch ( 1994 ) . Since then he has appeared in numerous films in his native Scandinavia and Europe in general , including Headhunters ( 2011 ) and A Thousand Times Good Night ( 2013 ) . In the United States , his debut film role was in the war film Black Hawk Down ( 2001 ) , playing Medal of Honor recipient Gary Gordon . He then played Detective John Amsterdam in the short-lived Fox television series New Amsterdam ( 2008 ) , as well as appearing as Frank Pike in the 2009 Fox television film Virtuality , originally intended as a pilot . He became widely known to a broad audience for his current role as Ser Jaime Lannister , in the HBO series Game of Thrones . In 2017 , he bec...	`Nukaaka Coster-Waldau Nukaaka Coster-Waldau ( née Sascha Nukaka Motzfeldt ; born 23 February 1971 ) , Nukâka , is a Greenlandic singer , actress , and a former Miss Greenland . She is married to Danish actor Nikolaj Coster-Waldau and has two daughters with him . The Greenlandic actress of Inuit , German and Norwegian extraction was born in Uummannaq , Greenland . She is the daughter of Vivi and Josef Motzfeldt , a Greenlandic politician .`	`Fox Channel (Germany) Fox ( also known as Fox Channel or FOX Channel Germany ) is a German pay-tv-channel dedicated to television drama series . The channel heavily features American and British drama series from the major broadcast networks and premium cable networks . They are broadcast without interruptions with both German and English audio .`	Fox Broadcasting Company The Fox Broadcasting Company ( often shortened to Fox and stylized as FOX ) is an American English language commercial broadcast television network that is owned by the Fox Entertainment Group subsidiary of 21st Century Fox . The network is headquartered at the 20th Century Fox studio lot on Pico Boulevard in the Century City section of Los Angeles , with additional major offices and production facilities at the Fox Television Center in nearby West Los Angeles and the Fox Broadcasting Center in the Yorkville neighborhood of Manhattan , New York City . It is the third largest major television network in the world based on total revenues , assets , and international coverage . Launched on October 9 , 1986 as a competitor to the Big Three television networks ( ABC , NBC and CBS ) , Fox went on to become the most successful attempt at a fourth television network . It was the highest-rated broadcast network in the 18 -- 49 demographic from 2004 to 2012 , and earne...	New Amsterdam (TV series) New Amsterdam is an American television drama which aired for eight episodes in 2008 on Fox . The series starred Nikolaj Coster-Waldau as `` John Amsterdam '' ( real name Johann van der Zee ) , an immortal Dutch man born in 1607 , who has lived in New York City on and off since he was 14 years old , and who is a homicide detective in the present day . The series was nominated for an Emmy for Main Title Design .	`Nick Coster Nick Coster ( born 24 September 1985 in Alphen aan den Rijn ) is a Dutch footballer who currently plays for Hoofdklasse side DOVO . He last played professionally as a left back for RKC Waalwijk in the Dutch Eerste Divisie , whom he had joined in 2014 .`	`Nicolas Coster Nicolas Dwynn Coster ( born December 3 , 1933 ) is a British-born American actor , most known for his work in daytime drama and as a character actor on nighttime television series , such as Buck Rogers in the 25th Century and Star Trek : The Next Generation .`	`Claus Spahn Claus Spahn ( born 15 May 1940 in Bottrop , Germany ) is a former TV programme editor of Germany 's Westdeutscher Rundfunk ( WDR ) broadcasting corporation , a presenter , producer and author . He was awarded the Cross of the Order of Merit of the Federal Republic of Germany in 2001 .`	Jonathan Elias Weiske Jonathan-Elias Weiske ( born in Berlin on 4 February 1996 ) is a German film and television actor and voice actor . Jonathan was born in Berlin son of Oliver Betke and Claudia Weiske , an artistic family , with his father being a film/television/theatre actor who had also appeared in musicals in addition to working also a synchronspeecher ( voice over ) and a photographer . Jonathan 's mother also worked as an actress who had appeared amongst others in Gute Zeiten , schlechte Zeiten in the role Elke Opitz . She had also appeared in a number of theater productions in Berlin ( in Theater am Kurfürstendamm ) , Dresden and in Hamburg . Jonathan started as a child actor at age five . He had his screen debut appearing in the series Herzschlag-Die Retterand two years later in the short film Jesus Christ , directed by U. Preatel . He became well-known through the series Fünf Sterne that ran from 2005 to 2008 on ZDF , where he starting as a 9-year old , he played the c...	Fox (Finland) Fox ( formerly SuomiTV ) is a Finnish entertainment TV channel owned and produced by Fox Networks Group ( FNG ) Nordic . The channel was acquired by FNG Nordic in January 2012 , and then relaunched as the current incarnation on 16 April that year . Fox is freely available in Finland both over-the-air and through cable , as it bases its funding on television advertisements . It is FIC Nordic 's first channel that does not operate on subscriber fees . The channel provides a free catch up service through its online platform FOXPlay	Fox News Fox News ( officially known as the Fox News Channel ) is an American basic cable and satellite news entertainment television channel owned by the Fox Entertainment Group , a subsidiary of 21st Century Fox . The channel broadcasts primarily from studios at 1211 Avenue of the Americas , New York City , New York . The channel was created by Australian-American media mogul Rupert Murdoch , who hired former Republican Party media consultant and CNBC executive Roger Ailes as its founding CEO . It launched on October 7 , 1996 , to 17 million cable subscribers . Fox News grew during the late 1990s and 2000s to become a dominant cable news network in the United States . Rupert Murdoch is the current chairman and acting CEO of Fox News . Fox News Channel has been accused of biased reporting and promoting the Republican Party . Critics have cited the channel as detrimental to the integrity of news overall . Fox News employees have responded that news reporting operates independentl...	Prva Srpska Televizija Prva ( -LSB- pr̩̂ːʋaː -RSB- ) or Prva Srpska Televizija ( Прва српска телевизија , in English : First Serbian Television ) is a Serbian commercial television network with national coverage . Launched on 31 December 2006 at 7pm as Fox televizija , the network has since December 2009 been owned by the Greek media company Antenna Group . From its inception until December 2009 , it was majority owned by the global media conglomerate News Corporation , which attached its Fox brand to the operation , naming Dan Bates as the CEO . During its time under News Corporation , Fox televizija closely co-operated with News Corporation owned networks in neighboring countries such as bTV in Bulgaria and Fox Turkey , sharing production capabilities on specific projects and making use of their experiences . Though quickly establishing itself on the Serbian television market and improving its viewership stake year after year , Fox televizija was a money-losing operation , postin...
`Roman Atwood is a content creator.`	Roman Atwood Roman Bernard Atwood ( born May 28 , 1983 ) is an American YouTube personality , comedian , vlogger and pranker . He is best known for his vlogs , where he posts updates about his life on a daily basis . His vlogging channel , [object Object] RomanAtwood '' , where he posts pranks . His prank videos have gained over 1.4 billion views and 10.3 million subscribers . Both of these channels are in the top 100 most subscribed on YouTube , and he became the second YouTuber after Germán Garmendia to receive two Diamond Play Buttons for his two channels .	`Jeff Atwood Jeff Atwood is an American software developer , author , blogger , and entrepreneur . He writes the computer programming blog Coding Horror . He co-founded the computer programming question-and-answer website Stack Overflow and co-founded Stack Exchange , which extends Stack Overflow 's question-and-answer model to subjects other than programming . Atwood 's most recent project as of 2012 is the development of Discourse , an open source Internet discussion platform .`	Brett Atwood Brett Atwood is a website editor , content strategist and former print and online journalist whose writings have appeared in Billboard , Rolling Stone , Vibe , The Hollywood Reporter and other publications . Atwood has held managing editor positions at leading Internet sites Amazon.com , Secondlife.com and RealNetworks . In 2001 , Atwood 's career shifted to include academia . He is currently the director of the Integrated Strategic Communication at Everett ( ISC ) program and an associate professor at the Edward R. Murrow College of Communication at Washington State University . He holds an M.A. degree in communication from University of the Pacific in Stockton , California .	`Roman Balleza Roman Balleza ( born March 18 , 1984 ) , better known as Roman X , is a Los Angeles -- based songwriter , producer , musician , and DJ . He and Dan Vikta make up the record production duo Rockystronic , producing pop , hip hop , rock and electronic music .`	`Dave Roman Dave Roman is an American writer and artist of webcomics and comics .`	`Roman Kemp Roman Kemp ( born 28 January 1993 ) is an American -- born British television and radio presenter . He is best known for working on the Capital FM radio station . He was also the social media correspondent for The X Factor UK 2016 .`	`Brian Atwood Brian Atwood ( August 4 ) is an American shoe designer . He studied at the Fashion Institute of Technology before becoming a model . He gave Versace management a handwritten resume which led to a job offer .`	`David Attwood David Attwood ( born 28 August 1952 ) is an English filmmaker .`	`Hugh Atkin Hugh Atkin is an Australian lawyer and notable producer of Internet viral videos . Atkin began producing videos for YouTube in 2007 , and his videos have appeared on the Australian Broadcasting Corporation 's Unleashed website .`	`Ryan Atwood Ryan Francis Atwood is a fictional character on the FOX television series The O.C. , portrayed by Benjamin McKenzie . A troubled teenager from Chino , California , he is the main protagonist of the show , an outcast who is given a second chance when the wealthy Cohen family takes him in .`	`Art Atwood Art Atwood ( December 27 , 1973 -- September 11 , 2011 ) was an American IFBB professional heavyweight bodybuilder .`	`John L. Roman John L. Roman is an American television and film producer . His major credits include Chicago Fire ( 2012-2014 ) , Law and Order : Criminal Intent ( 2001-2010 ) and Deadline ( 2000-2001 ) .`

Loss: pylate.losses.contrastive.Contrastive

msmarco_distillation_simlm_rescored_reranked_min15

Dataset: msmarco_distillation_simlm_rescored_reranked_min15
Size: 485,721 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 5 tokens mean: 10.4 tokens max: 25 tokens	min: 16 tokens mean: 38.77 tokens max: 39 tokens	min: 23 tokens mean: 38.92 tokens max: 39 tokens	min: 21 tokens mean: 38.8 tokens max: 39 tokens	min: 24 tokens mean: 38.85 tokens max: 39 tokens	min: 17 tokens mean: 38.8 tokens max: 39 tokens	min: 22 tokens mean: 38.85 tokens max: 39 tokens	min: 19 tokens mean: 38.9 tokens max: 39 tokens	min: 23 tokens mean: 38.86 tokens max: 39 tokens	min: 17 tokens mean: 38.85 tokens max: 39 tokens	min: 21 tokens mean: 38.83 tokens max: 39 tokens	min: 17 tokens mean: 38.85 tokens max: 39 tokens	min: 19 tokens mean: 38.89 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`what temperature does brass melt`	Brass has a relatively low melting point compared to iron, steel, or gold, but it still requires a special furnace. Many beginning metalworking hobbyists start out with aluminum, which is easier to melt, but brass is often the next step.ost brass will melt at temperatures as low as 1,650ºF (900ºC), but a higher maximum temperature will give you a margin of error, and make the brass easier to pour. 1 Select a furnace large enough to hold the crucible and brass you intend to melt. 2 Consider the fuel you will be using. 3 3 Edit step.	`Brass has higher malleability than bronze or zinc. The relatively low melting point of brass (900 to 940 °C, 1652 to 1724 °F, depending on composition) and its flow characteristics make it a relatively easy material to cast.`	`Melting aluminum, brass, or other mid-temperature melting point metals requires over 1000 degree temperatures.To build a simple home made furnace for casting metals you need to meet some simple design requirements and use appropriate materials.elting aluminum, brass, or other mid-temperature melting point metals requires over 1000 degree temperatures.`	`Not all brass is made the same way. The melting point of red brass is 1,832 degrees Fahrenheit. The melting point of yellow brass is 1,660-1,710 degrees Fahrenheit, and th … e melting point of admiralty brass is 1,650-1,720 degrees Fahrenheit.`	`Brass has higher malleability than bronze or zinc. The relatively low melting point of brass (900 to 940 °C, 1652 to 1724 °F, depending on composition) and its flow characteristics make it a relatively easy material to cast.o enhance the machinability of brass, lead is often added in concentrations of around 2%. Since lead has a lower melting point than the other constituents of the brass, it tends to migrate towards the grain boundaries in the form of globules as it cools from casting.`	`Compared to iron's melting point -- 2,786 degrees Fahrenheit -- the average melting point of brass at 1,688 makes it a low-melting alloy.`	Three Parts: Setting up a furnace Following safe practices Melting the brass. Brass has a relatively low melting point compared to iron, steel, or gold, but it still requires a special furnace. Many beginning metalworking hobbyists start out with aluminum, which is easier to melt, but brass is often the next step.ost brass will melt at temperatures as low as 1,650ºF (900ºC), but a higher maximum temperature will give you a margin of error, and make the brass easier to pour. 1 Select a furnace large enough to hold the crucible and brass you intend to melt. 2 Consider the fuel you will be using.	`Melting the brass. 1 1. Heat your molds and skimmer spoon. Preheat the metal molds to a temperature above 212ºF (100ºC) to remove all moisture, or the molten metal will spatter when poured. 2 2. Place your crucible in your furnace. Position the crucible into your furnace. 3 3. Light the furnace.`	`By varying the proportions of copper and zinc, the properties of the brass can be changed, allowing hard and soft brasses. The relatively low melting point (900-940 °C depending on composition) of brass and its flow characteristics make it a relatively easy material to cast.`	`Put all of the scraps of one type into the crucible, and apply a layer of baking soda over the metal. Raise the flame on the torch according to the type of metal you need to melt. Copper, brass and gold will all melt between 1700 and 2000 degrees, while iron and steel will not melt until nearly 2800 degrees. Put on gloves and apply the torch flame to the crucible and wait for the metal to melt completely into liquid.`	`Brass has higher malleability than zinc or copper. It has a a low melting point (900 centigrade) and flows when melted making it easy to cast in molds. Combinations of iron, aluminum, silicon and manganese make brass wear and tear and corrosion resistant. Susceptible to stress cracking when exposed to ammonia.`	Thus if the plumber were to add aluminium, instead of tin, to the lead, the two metals would not dissolve - they would behave like oil and water. When cast, the metals would separate into two layers, the heavy lead below and aluminium above. One difficulty in making alloys is that metals have different melting points. Thus copper melts at 1,083C, while zinc melts at 419C and boils at 907C So, in making brass, if we just put pieces of copper and zinc in a crucible and heated them above 1,083C, both the metals would certainly melt.
`who developed surgical techniques to prevent the entry of microorganisms into wounds`	`Joseph Lister, 1st Baron Lister, OM PC PRS, known between 1883 and 1897 as Sir Joseph Lister, Bt., was a British surgeon and a pioneer of antiseptic surgery. He promoted the idea of sterile surgery while working at the Glasgow Royal Infirmary. Lister successfully introduced carbolic acid to sterilise surgical instruments and to clean wounds. Applying Louis Pasteur's advances in microbiology, Lister championed the use of carbolic acid as an antiseptic, so that it became the first widely used anti`	`Aseptic technique is a method used to prevent microbial contamination of wounds and. other susceptible sites by ensuring that only sterile objects and fluids are used. Aseptic. technique reduces the risk of contamination to vulnerable sites from organisms that. could cause infection (Fraise and Bradley, 2009).`	`Joseph Jackson Lister invented the achromatic microscope in 1826. His son, Joseph Lister was a surgeon who began the idea of antiseptic surgery. He wanted to kill microbes tha … t caused infections such as gangrene in wounds.`	`Techniques that prevent the entry of microorganisms into sterile tissues Asepsis Examples Cleansing the skin with iodine prior to surgery; using sterile needles`	`In 1941, a doctor, Charles Fletcher, at a hospital in Oxford had heard of their work. He had a patient who was near to death as a result of bacteria getting into a wound. Fletcher used some of Chain’s and Florey’s penicillin on the patient and the wound made a spectacular recovery. Unfortunately, Fletcher did not have enough penicillin to fully rid the patient’s body of bacteria and he died a few weeks later as the bacteria took a hold.`	`surgical asepsis the exclusion of all microorganisms before they can enter an open surgical wound or contaminate a sterile field during surgery. See accompanying table. Measures taken include sterilization of all instruments, drapes, and all other inanimate objects that may come in contact with the surgical wound.`	Asepsis is the absence of pathogenic microorganisms that cause disease. It can also be referred to as clean technique (Phillips, 2013). Surgical aseptic technique is a method employed to maintain asepsis and minimise the risk of introducing pathogens into a surgical wound (Rowley et al, 2010).sepsis is the absence of pathogenic microorganisms that cause disease. It can also be referred to as clean technique (Phillips, 2013). Surgical aseptic technique is a method employed to maintain asepsis and minimise the risk of introducing pathogens into a surgical wound (Rowley et al, 2010).	`As the germ theory of disease became more understood, it was realised that infection could be better avoided by preventing bacteria from getting into wounds in the first place. This led to the rise of less primitive surgery.`	`This is achieved by isolation precautions, in which the objects in the patient's environment are protected from contamination or disinfected as soon as possible after contamination. surgical asepsis the exclusion of all microorganisms before they can enter an open surgical wound or contaminate a sterile field during surgery. See accompanying table. Measures taken include sterilization of all instruments, drapes, and all other inanimate objects that may come in contact with the surgical wound.`	`Bayliss Bay·liss (bā'lĭs), Sir William Maddock. 1860-1924. British physiologist. With Ernest Starling he discovered (1902) secretin. He also developed a treatment for surgical or wound shock in which saline injections replaced lost blood, a technique credited with saving many lives in World War I.`	`Along with the invention of antiseptics by the English surgeon Joseph Lister, penicillin was used in World War II to keep infections from spreading. By attacking open wounds with penicillin and sterilizing surgeries with antiseptics, army doctors were able to revitalize troops more quickly.`	`A surgical mask, also known as a procedure mask, is intended to be worn by health professionals during surgery and during nursing to catch the bacteria shed in liquid droplets and aerosols from the wearer's mouth and nose. Its first recorded use was by the French surgeon Paul Berger during an 1897 operation in Paris.`
`what shots do puppies need`	`Distemper, parvovirus, hepatitis and rabies are designated “core” vaccinations. All puppies should receive these, and often they’ll be combined into a single shot. “Noncore” vaccinations are recommended for puppies at specific risk for those conditions, such as Lyme disease, kennel cough and leptospirosis.`	Most puppies need at least three DAPP vaccines, two Bordetella vaccines, and one rabies vaccine during these visits to be protected for a full year. The next vaccine visit should occur approximately one year after the last puppy shots were given.ther vaccines are available but should only be considered under special circumstances. Often, protection against several diseases is offered in a single inoculation. For example, a DAPP vaccine protects against distemper, adenovirus, parvovirus and parainfluenza.	`All puppies need to have the core shots to protect them from the most common dog illnesses. But there are also other diseases/conditions that some pups may be at risk of catching. These can also be prevented by vaccination, but it's important to know that not all puppies need (or should have) these.`	`Puppies Need Parvo Vaccines. Parvo vaccines typically begin between 6 and 8 weeks of age, with boosters every three to four weeks until the puppy is 16 to 20 weeks old. An additional booster is administered at the puppy’s one-year checkup.`	`When do puppies get their first shots? A: Puppies should get their first combination vaccine shows when they are between six and seven weeks old. These shots include vaccines for distemper, hepatitis, coronavirus, parainfluenze and paraovirus.`	`Rabies shots are one of the vaccines you will need to get your new puppy. These shots can be given to puppies that are 16 weeks of age. You will also need to get a vaccine booster when your puppy is a year old. Additional boosters need to be given every three years after the first two Rabies vaccines.`	These will be the DHLPP (Distemper, Hepatitis, Leptospirosis, Parvovirus and Parainfluenza) in addition to the Corona shot. Your puppy will need to return to the vet twice more for the same shots; once between 9 and 11 weeks and then again sometime between 12 and 14 weeks.During the last visit your puppy will also be immunized against Rabies. After these initial shots are completed then the suggested return for boosters in once per year.our puppy will need to return to the vet twice more for the same shots; once between 9 and 11 weeks and then again sometime between 12 and 14 weeks. During the last visit your puppy will also be immunized against Rabies. After these initial shots are completed then the suggested return for boosters in once per year.	`You can use this chart to help you keep track of when your puppy needs which shots: DHLPPC. First vaccination: 6 to 8 weeks; Second vaccination: 9 to 11 weeks; Third vaccination: 12 to 14 weeks; Fourth vaccination: 16 to 17 weeks; Booster shots: 12 months; Bordetella. First vaccination: 14 weeks; Booster shots: 6 months; Rabies`	`Rabies Shots Are Essential. Rabies shots are one of the vaccines you will need to get your new puppy. These shots can be given to puppies that are 16 weeks of age. You will also need to get a vaccine booster when your puppy is a year old. Additional boosters need to be given every three years after the first two Rabies vaccines.`	Rabies Shots Are Essential. Rabies shots are one of the vaccines you will need to get your new puppy. These shots can be given to puppies that are 16 weeks of age. You will also need to get a vaccine booster when your puppy is a year old.Additional boosters need to be given every three years after the first two Rabies vaccines.ou're going to be firm and tell the vet that you don't need a bunch of shots when you go in for that appointment in two weeks-just the Rabies vaccine. Actually, distemper is quite serious. While it is not dangerous to all the people in your household, the way Rabies is, it is usually fatal in puppies.	`Puppy’s first shots. Puppies need their first vaccinations at six to eight weeks old. Whether you adopt a puppy or buy one, make sure you get any medical records. In their first year, fur-babies will need to visit the vet a few times for immunizations and regular booster shots.`	`Distemper Shots. Your vet says your puppy needs a whole series of distemper shots. He or she gives your puppy the first shot, gives you a schedule of vaccines and sets up the next appointment, which is two weeks away.`

Loss: pylate.losses.contrastive.Contrastive

reddit_triples

Dataset: reddit_triples
Size: 200,000 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 9 tokens mean: 30.8 tokens max: 39 tokens	min: 8 tokens mean: 29.2 tokens max: 39 tokens	min: 9 tokens mean: 29.55 tokens max: 39 tokens	min: 9 tokens mean: 29.51 tokens max: 39 tokens	min: 9 tokens mean: 30.05 tokens max: 39 tokens	min: 9 tokens mean: 29.58 tokens max: 39 tokens	min: 8 tokens mean: 30.01 tokens max: 39 tokens	min: 8 tokens mean: 29.48 tokens max: 39 tokens	min: 9 tokens mean: 29.82 tokens max: 39 tokens	min: 9 tokens mean: 29.54 tokens max: 39 tokens	min: 10 tokens mean: 30.1 tokens max: 39 tokens	min: 9 tokens mean: 29.58 tokens max: 39 tokens	min: 9 tokens mean: 29.92 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`A palace coup, not a total revolution. Mere change in leadership, not an upending of society.`	I don't understand why you guys seem to think revolutions only occur if poor people are now the ones in charge.[object Object][object Object]Did that even happen in the French Revolution? Were they too not led by their own social elite? The same class of people in fact who led the American Revolution. There is nobility and there is everyone else. Within the latter category there are different classes. The higher end of the common class is behind both the French and American revolutions. And all common people fought in both revolutions. [object Object][object Object]The French was more severe, more loud, and overall more emphatic for many reasons described elsewhere in this thread. It took place IN the actual country, the King of executed, the monarchy was destroyed literally lol. I get all that. [object Object][object Object]But really for some reason there is this repeated rewriting of history going on here in regards to the American Revolution. And I get it Americans love to glamorize their past so there is an impulse among historians or just history buffs in genera...	`That makes a lot of sense. Camilla would have joined you as a Minerva if Conquest didn't exist, Leo seems like he would be your Gotoh.`	`I would say:[object Object][object Object]-The Old Hunters[object Object][object Object]-Artorias of the Abyss[object Object][object Object]-Blood Dragon[object Object][object Object]-Red Dead Redemption: Undead Nightmare[object Object][object Object]-Far Cry 4: Yeti Valley[object Object][object Object]All great DLC's.[object Object]Wasn't much a fan of The Witcher, personally.`	`He was great in the Hall of Fame ceremony as well. Also huge props to the technicians and camera crew because they work like hell. edit: He lowkey roasted JBL when he said "DDP Yoga worked for me" JBL has a Wrestling God physique`	`Ah I see. I was misremembering then`	`1960. People want to act like discrimination and oppression only occurred during slavery and in 1960 a security detail was needed for a little girl to go to school.`	`Don't you know that all disabilities become genetic and also are transmittable?[object Object][object Object]I once knew someone whose idiot spouse didn't want their kid to play with its cousin because the kid might catch autism from the cousin.`	`Those monsters give the rest of us vegans a bad name. I can't stand them. There is so much anti-protein anti-fat stuff going on in the vegan community right now, which is ridiculous because there are plenty of ways to get both. I'll admit, they had me going for a while. It's only in the past month that I've started eating olive oil again. And it was scary as shit...I was convinced that 2tsps of olive oil was going to give me insta heart disease.`	Labyrinth for [object Object] (she's 4 years older than me, so 12 year old me crushing on 16 year old her was A-OK)[object Object][object Object]Legend for [object Object].[object Object][object Object]Flight of the Navigator for [object Object]. Go on, throw rotten fruit at me :P[object Object][object Object]The Last Starfighter for [object Object].[object Object][object Object]Can I sneak in a 1991 movie? The Silence of the Lambs for [object Object] (she was the senator's daughter who got abducted by buffalo bill). (yes, that's a weird one, I make no excuses or apologies for it)	Dude maybe you are my long lost sibling because that’s also my dad. The most important thing he will point out when I counter his propaganda with facts and evidence is that I should start thinking for myself like he does. [object Object][object Object]This mindset starts with religious indoctrination. If they can get you to believe in talking snakes, an instant global flood, or a dude who walks on water then they can get you to believe anything. It’s as if they want adults with the mind of a child that they can easily manipulate.	`I feel like Oscar from the office was extremely flanderized in the later seasons. In the beginning he's a very normal, rational guy who happened to be gay. By the end he's doing ridiculous things driven by his sexuality. Most office background characters experienced some degree of flanderization, as do most comedy characters now that I think about it.`	`Lmao, [object Object].[object Object][object Object]Everyone is a Nazi but the guys threatening to bust heads unless they get their way. Note, the linked individual is a reporter for Washington Monthly and a longtime democratic activist.`
They speak of war, but they are the ones holding guns and shooting at defenceless people [object Object][object Object][object Object]I suspect, like the terrorists, they are looking for a reaction. Perhaps a ddos of core in retaliation or at least of bitcointalk so as to justify their shooting at innocent people. I'm not smart enough to know of the best response. Ddosing core nodes is terrible pr. Doing nothing has no deterrence capability. [object Object][object Object]But then, it isn't doing nothing. It is in fact doing a lot. Putting up with the inconvenience, perhaps even finding better ddos protection, withstanding the assault and standing firm. Showing to those who wish to dictate that we do not need to return bullets, we work with code.[object Object][object Object]	Everything happening is showing exactly what the best response is. These apply to all Bitcoiners.[object Object][object Object]* Run a node. It is a challenge to flood thousands of nodes at once. If they manage to succeed DoS in that environment, it is likely a flaw and can be fixed.[object Object][object Object]* Do not mine on centralized services. This should have been fixed long ago, but at least p2pool exists. More demand can translate to more user friendly solutions.[object Object][object Object]* Use censorship-resistant channels to communicate. Same as above, more people making an effort means more future possibilities. [object Object][object Object][object Object]* Put an effort into testing/utilizing existing solutions (such as Lighthouse?).[object Object][object Object]	`Actually gonna miss LeBlanc's global ult clone[object Object][object Object]Idk if it was actually any good but the concept of it was hilarious`	`Come ho fatto a perdermelo? Mannaggia.[object Object][object Object]A saperlo l'avrei registrato su VHS.`	`Two ways to do that, of course. You can just never play C-game heroes or you can improve your C-game to A-game. Takes longer, but I do feel like I'd rather get more out of the game than just a number. This should really be "how to climb faster," not "how to climb, period."`	`You really are the worst, u/bread_heads. THE WORST!`	`Was pretty funny how he just walked in and started and none of the media was there yet.`	`> I wish people didn't get so divided by the #TeamPaxton vs. #TeamTrevor nonsense...[object Object][object Object]Like the FO. :)`	`oh sorry my phone kinda sucks and i guess each time i clicked add comment it added one whoops i just thought it wasnt working at all`	Okay, so there are like two/three situations that can be happening here. [object Object][object Object]1. Your friend is jealous of you for some reason. Maybe you're more interesting or prettier or something. [object Object][object Object]2. You're genuinely awkward, and she's trying to help you out. [object Object][object Object][object Object]3. Maybe you embarrass her or something. Who even knows why. [object Object][object Object]Why don't you talk to her and explain that you don't think you're being awkward. Also, mention she's hurting your feelings and making you feel weird around people by saying this stuff to you. You definitely have to talk to her though.	`Also, take me to the page I clicked on not the fucking home page.`	`Hey son! Want to go on a hunting trip?`	`Why do engineers and professors end up working at uber or a small office? Shit happens.`
`one best friend is an airhead, the other has her shit together.`	`Re: your flair.... god I hate Thursdays. They're just far enough away from Monday that you don't have any more momentum left over and they embody the disappointment that is every day that isn't Friday by being not quite Friday. They're the doldrums of the week and the most useless day.`	`When she first walks in with the torn-up outfit...[object Object][object Object]Mommy`	Shouldn't morty have a passive ability that Rick has hexproof/shroud?[object Object][object Object]Also, rick should be more like a stone forge mystic, pulling artifacts out of the deck instead of destroying them.[object Object][object Object]I love Jerry, completely useless. [object Object][object Object]Summer, is nice, what if she forced banding between Beth and Jerry? [object Object][object Object]The Meeseeks concept is good. You would have to exile the spell and have the ability allow you to cast it from exile; in order to be able to have the middle ability work properly. Also, for flavor I would probably have Meeseeks kill you if you haven't cast the spell in a certain amount of turns. Something like, "During your upkeep put a time counter on ~. At the end of your turn if there are 3 or more time counters on ~ you lose the game."[object Object][object Object]	`Feel bad for the Canadian goalie. Media is going to be all over the poor kid.`	`[object Object]`	`"Not a fan of putting my hands down there to begin with" says it all, really.`	`I just learned in one of these threads that he's still holding office and quite popular regionally! From the brief googling and skimming I did he's been doing a real nice job of representing his district and seems to have mostly been forgiven... after all, what 45 year old doesn't dream of running away with an Argentinian woman once in a while?[object Object][object Object]Edit: To be clear, I'm not saying what he did was ok. Just that I understand why the voters forgave him.`	They used to be friends. The happy days ended in 2012 when Mayweather was sentenced to 90 days in prison for assaulting a woman, the mother of one of his four children. In the fighter’s absence, 50 Cent was accused of interfering with Mayweather’s team of boxers and the friendship ended.[object Object] 50 Cent still claims that Money owes him $2 million.[object Object]He made the abovementioned challenge to humiliate Mayweather. In response, Mayweather said : [object Object]“Making fun out of a person because he can’t read is not funny – it is tragic, I would be perfect at reading if it was how I made my living and fed my family. I’m a boxer, that’s what I do and I will say this: intelligence and education are two totally different things.”[object Object][object Object]That was perfect imo.	`So your religion teacher lied to you?[object Object][object Object]That's not setting good morals`	I think it's the simplistic grindy nature of it. No real story, no handholding, first iteration had very little variety. However above all else it really nailed the effort/reward equilibrium. Many games fail to make you feel accomplished without pandering to you or making you overpowered.[object Object][object Object]They kept that idea in mind with each game but improved just about every other system along the way. What was a niche series growing up is now in the spotlight and I'm glad it's getting the recognition it deserves in this day of age of greedy MTX/cash grabs.	> As a new grad in a project management position, you will learn an immense about about how the organization works. However, you might never do any design or analysis of the sort you learned in college.[object Object][object Object]Yep, and the most common impact I've seen from this is not knowing the impact of problems. Someone who hasn't done any design doesn't have much in the way of critical judgement on whether one problem is worse than another, which means they are reliant on accurate information from their engineers. Not an issue provided your engineers a) are good at explaining things b) aren't motivated to cover up their own mistakes...	`Also I feel like there are tons of people who still don’t like Jeffree, it’s not like he’s some beloved pillar of the community. He has his fans, just like Laura has hers. It just so happens that his offenses aren’t as fresh.`

Loss: pylate.losses.contrastive.Contrastive

medi_sts_flickr_sampled

Dataset: medi_sts_flickr_sampled
Size: 51,186 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 8 tokens mean: 16.22 tokens max: 39 tokens	min: 6 tokens mean: 16.34 tokens max: 39 tokens	min: 8 tokens mean: 16.96 tokens max: 39 tokens	min: 8 tokens mean: 15.88 tokens max: 39 tokens	min: 7 tokens mean: 16.18 tokens max: 39 tokens	min: 8 tokens mean: 16.02 tokens max: 39 tokens	min: 9 tokens mean: 16.03 tokens max: 36 tokens	min: 6 tokens mean: 16.07 tokens max: 39 tokens	min: 9 tokens mean: 16.07 tokens max: 39 tokens	min: 7 tokens mean: 16.17 tokens max: 39 tokens	min: 8 tokens mean: 16.24 tokens max: 39 tokens	min: 7 tokens mean: 16.11 tokens max: 39 tokens	min: 8 tokens mean: 16.3 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`A man is looking in a row of windows outside by a trash dumpster.`	`A man enjoys a cold beer and a cigarette in an empty lot.`	`Young girl in a blue jacket holding young boy in an orange shirt.`	`A man is holding a tennis racket getting ready to hit a ball.`	`a bear sitting in the corner of a habitat just sitting there`	`A giraffe is standing in the grass under a palm tree.`	`A small orange airplane is parked by a building.`	`A brown bench sits in the middle of a park.`	`A giraffe rubbing it's nose on a tree trunk.`	`A woman wearing sunglasses holds two newspapers.`	`A girl wearing camouflage pants sits on top of a Hummer.`	`Group of ducks swimming in a large body of water.`	`a man and woman wearing khaki smiling.`
`A man with a tan jacket is walking past a woman in a ponytail.`	`A man in a suit is walking towards a redheaded woman with her hair tied up.`	`Three boys diving into a lake.`	`Four pieces of luggage in a living room.`	`Three children are running down a grassy hill.`	`a man and woman sits in front of a pizza`	`A black and white photo of two woman singing under umbrellas.`	`A man is holding a bunch of banana's`	`A boy doing a trick on his skate board.`	`A man with a blue helmet is on a motorcycle on a blue skied day.`	`A man riding a skateboard with a helmet on in between cones.`	`A skateboarder who is jumping high into the air.`	`A bus stop with a couple of buses and people.`
`A dog makes it way through tall weeds.`	`A beaver on the shore of a stream.`	`A man jumps into the air while a young girl pretends to fight back while standing on a couch.`	`A young child taking a swing at a baseball`	`a living room with several chairs and a small table`	`A boy holding a large foam wand waves it near a swimming pool.`	`a man on a surfboard jumping over a wave`	`A man in a green shirt is skating on a railing.`	`A hodge podge of colors and patterns decorate a bathroom.`	`a white dog jumping towards the camera`	`A group of fourteen men posing for a picture together.`	`A man in a black dress is sitting and sleeping in a train.`	`Young foreign woman with bracelets on taking a picture of something in front of her with a bus behind here`

Loss: pylate.losses.contrastive.Contrastive

nq_cocondensor_hn_mine_reranked_min15

Dataset: nq_cocondensor_hn_mine_reranked_min15
Size: 70,002 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 11 tokens mean: 13.49 tokens max: 24 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens	min: 39 tokens mean: 39.0 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`big little lies season 2 how many episodes`	Big Little Lies (TV series) series garnered several accolades. It received 16 Emmy Award nominations and won eight, including Outstanding Limited Series and acting awards for Kidman, Skarsgård, and Dern. The trio also won Golden Globe Awards in addition to a Golden Globe Award for Best Miniseries or Television Film win for the series. Kidman and Skarsgård also received Screen Actors Guild Awards for their performances. Despite originally being billed as a miniseries, HBO renewed the series for a second season. Production on the second season began in March 2018 and is set to premiere in 2019. All seven episodes are being written by Kelley	Big Love seasons, the song "Home", performed by the British band Engineers, was adopted as the show's theme song along with a new title sequence. "God Only Knows" was covered by Natalie Maines for the closing of the series finale. David Byrne recorded a complete soundtrack to the second season, released as "" on August 19, 2008. Review aggregate Metacritic indicated positive critical response for all five seasons. The average scores for the first through fourth seasons were 72/100, 71/100, 79/100, and 70/100, indicating "generally favorable reviews." The fifth and final season received an average score of 85/100, or "universal acclaim". Based	Big Love and Margie rush to his aid, and perform a blessing over him as he bleeds to death on the ground. Eleven months later, a memorial is held in Bill's honor, and Sarah returns home with Scott and her newborn son, who they've named after Bill. As Sarah and Scott prepare to fly back to Oregon, Margie also prepares to leave on her mission to South America. The wives emotionally embrace as Bill watches from the dining table. The series' theme song for its first three seasons was "God Only Knows" by The Beach Boys. For the series' fourth and fifth	Little Fires Everywhere (miniseries) Street has an overall deal. On March 2, 2018, the production was officially announced but a network had yet to be determined. The series is set to be written and showrun by Tigelaar who will also executive produce alongside Witherspoon, Washington, Lauren Neustadter, and Pilar Savone. The author of the novel, Ng, will act as producer. Production companies involved in the series include Hello Sunshine, Simpson Street, ABC Signature Studios. On March 12, 2018, it was announced that Hulu had been given the production an eight episode order. This series order reportedly came after a multiple-outlet bidding war involving various	Pretty Little Liars (season 2) to 25 episodes. On December 13, 2011, it was announced that the identity of A would be revealed in the Spring Finale. Plus, the season finale will see someone arrested for the murder of Alison, as well as the death of a beloved character — excluding the four female leads. Filming of season two wrapped on December 16, 2011. Janel Parrish, Tammin Sursok, Bianca Lawson, and Tyler Blackburn return in the second season as Mona Vanderwaal, Jenna Marshall, Maya St. Germain, and Caleb Rivers. Also returning are Yani Gellman, Torrey DeVitto, Lindsey Shaw, Claire Holt, Keegan Allen, Brant Daugherty, Brendan	Secrets and Lies (U.S. TV series) to his family's private equity firm, when his wife Kate is murdered. The series was renewed for a second season on May 7, 2015, which premiered on September 25, 2016 and concluded on December 4, 2016. On May 11, 2017, ABC cancelled the series after two seasons. Ben Crawford is a self-employed contractor, married to Christy with whom he has two daughters, Natalie, 16, and Abby, 12. His best friend, Dave, lives in their summer house. They have a neighbor, Jess, who is estranged from her husband Scott, who is in the military. Jess and Scott have a five year	Little Big Shots (UK TV series) judges, winners or losers. In each episode, Dawn French talks to a 3–13 year old child about their talent, before the child performs it in front of a studio audience. Some talents shown include singing, dancing, cooking and karate. The first series began on 1 March and ended on 12 April 2017. It contained six episodes and was broadcast on Wednesdays. The second series of "Little Big Shots" began on April 22, 2018. As per the first series, this series will feature six episodes, but instead air on Sundays. The final episode will air on May 27, 2018, due to	Big Little Lies (novel) Lies", filmed on location in Monterey, California. It was originally released February 19, 2017 on HBO. The mini-series stars Nicole Kidman, Reese Witherspoon, Shailene Woodley, Alexander Skarsgård, Adam Scott, Laura Dern, and Zoë Kravitz, and won 8 Emmy Awards at the 2017 Primetime Emmy Awards including Outstanding Miniseries. Despite being called a miniseries, HBO announced in December 2017 that the series would return for a second season. Big Little Lies (novel) Big Little Lies is a 2014 novel written by Liane Moriarty. It was published in July 2014 by Penguin Publishing. The novel made the "New York Times" Best Seller	Little Big Shots (UK TV series) Little Big Shots (UK TV series) Little Big Shots (also known as Little Big Shots with Dawn French) is a British television talent show that premiered on ITV on 1 March 2017. The series is hosted by Dawn French and based on the American show of the same name created by Ellen DeGeneres and Steve Harvey who is also the host of the original. The series is produced by Wall to Wall. In June 2017, it was confirmed it would return for a second series, also with six episodes. The show, unlike shows such as "The X Factor", has no	Pretty Little Liars (season 2) girls learn of Mona's plan and rush to her aid. Spencer and Mona struggle, and Mona is knocked into a ravine, but survives. Emily learns that Maya was murdered, after she had seemingly disappeared. Mona is hospitalized in Radley Sanitarium, where she is visited by a blonde woman in a red trench coat, revealing Mona was working for someone else the whole time. [object Object] After an initial order of 24 episodes, it was announced in June that a special Halloween-themed flashback episode would air as part of ABC Family's 13 Nights of Halloween lineup, bringing the season 2 episode order	Big Fat Lie catchy chorus. The song peaked at number The accompanying music video was directed by Dawn Shadforth and sees Scherzinger frolicking at a beach in Malibu, California. Although BBC Radio 1 decided not to play it, the song entered the UK Singles Chart at number six and received moderate success in other international markets, peaking at number 22 in France, number 62 in Switzerland and number 99 in Germany. "Run" was released and impacted contemporary hit radio stations on September 30 as the second single worldwide and the lead single in the US. The third single, "On the Rocks", was released	Little Big Shots Little Big Shots Little Big Shots is an American variety television series. Co-created and produced by Steve Harvey and Ellen DeGeneres, with Harvey serving as host, the series features children demonstrating talents and engaging in conversation with Harvey. The series was ordered by NBC in May 2015 for an eight-episode first season, which premiered on March 8, 2016. On March 14, 2016, NBC renewed the series for a second season, which premiered on March 5, 2017. The series has since been renewed for third season, which premiered on March 18, 2018. The series premiered as a "preview" after "The Voice"
`who sang waiting for a girl like you`	Waiting for a Girl Like You Waiting for a Girl Like You "Waiting for a Girl Like You" is a 1981 power ballad by the British-American rock band Foreigner. The distinctive synthesizer theme was performed by the then-little-known Thomas Dolby, and this song also marked a major departure from their earlier singles because their previous singles were mid to upper tempo rock songs while this song was a softer love song with the energy of a power ballad. It was the second single released from the album "4" (1981) and was co-written by Lou Gramm and Mick Jones. It has become one of the band's most	Waiting for a Girl Like You held off the number 1 spot by Olivia Newton-John's single "Physical" for nine consecutive weeks, and then by Hall & Oates' "I Can't Go for That (No Can Do)" for a tenth week on January 30, 1982. Because of its chart longevity, it ended up being the number 19 song on the Top 100 singles of 1982. The song was the band's biggest hit until "I Want to Know What Love Is" hit number 1 in 1985. The song lists at number 100 on ""Billboard"'s Greatest Songs of All Time". Waiting for a Girl Like You "Waiting for a Girl	Girls Like You Girls Like You "Girls Like You" is a song recorded by American band Maroon 5, from their sixth studio album "Red Pill Blues" (2017). A second version featuring American rapper Cardi B was released by 222 and Interscope Records as the album's third single on May 30, 2018. This version was written by Adam Levine, Cirkut, Cardi B, Starrah, Jason Evigan, and Gian Stone and was produced by Cirkut and Evigan. The single reached number one on the US "Billboard" Hot 100 chart, making it Maroon 5's fourth and Cardi B's third chart-topper, who extended her record for most number-ones	Waiting for Love (Avicii song) Waiting for Love (Avicii song) "Waiting for Love" is a song by Swedish DJ and music producer Avicii, produced by Avicii and Dutch producer Martin Garrix and featuring uncredited vocals from Simon Aldred, the lead singer of former English band Cherry Ghost. The track was released on 22 May 2015 as the lead single from Avicii's second studio album, "Stories" (2015). The lyrics were also written by Aldred. The track was first premiered at Ultra Music Festival 2015 both by Avicii and Dutch electronic producer Martin Garrix. Prior to its release on May 22, 2015, it was thought that the	Girls Like Girls Like "Girls Like" is a song written and performed by British rapper Tinie Tempah and Swedish singer Zara Larsson. The song was released as a digital download in the United Kingdom on 26 February 2016 as the second single from Tempah's third studio album, "Youth". The song entered the UK Singles Chart at number 15, and it later peaked at number five, becoming Tempah's thirteenth top five single, and Zara Larsson's third. Tempah previewed a clip onto his Instagram account, showing him and two girls. The full music video premiered on his YouTube account on 17 March 2016. The	With a Girl Like You With a Girl Like You "With a Girl like You" is a song released by the English rock band the Troggs, written by Reg Presley and produced by Larry Page. The song reached number one in the UK Singles Chart on 4 August 1966, where it remained for two weeks. In the U.S., it peaked at #29 on the "Billboard" Hot 100. The song is featured, uncut, in a school dance scene from the 1991 Nicole Kidman/Noah Taylor film, "Flirting". It is also featured in the films and/or soundtracks to "Shine", "The Good Night" and "The Boat That Rocked". The	Waitin for You Waitin for You "Waitin for You" is a song recorded by American singer Demi Lovato featuring rapper Sirah. The duo along with Julia Michaels, Jason Evigan and Steve Mac composed the song for Lovato's fifth studio album "Confident" (2015). Evigan and Mac also handled the song's production, with Mitch Allan serving as a vocal producer. In December 2013, Lovato confirmed that she had started working on her then-untitled fifth studio album "Confident". The majority of the album's songs was recorded upon the conclusion of the Demi World Tour in 2015. "Waitin for You" was written by Lovato, Julia Michaels, Jason	Waitin for You Lovato's "most powerful" video yet. Lovato and Sirah performed the song live for the very first time on August 13, 2016 at the MGM Grand Garden Arena in Las Vegas during the Future Now Tour. Credits adapted from "Confident" album liner notes. Waitin for You "Waitin for You" is a song recorded by American singer Demi Lovato featuring rapper Sirah. The duo along with Julia Michaels, Jason Evigan and Steve Mac composed the song for Lovato's fifth studio album "Confident" (2015). Evigan and Mac also handled the song's production, with Mitch Allan serving as a vocal producer. In December 2013,	A Girl Like You (Easton Corbin song) of many a recent Nashville hit", and states that she is unlike any other girl. The song has sold 96,000 copies in the United States as of February 2018. A Girl Like You (Easton Corbin song) "A Girl Like You" is a song written by Ashley Gorley, Jesse Frasure, and Rhett Akins and recorded by American country music singer Easton Corbin. The song is Corbin's tenth single, and was originally intended to be the lead single from his upcoming fourth studio album before his termination from Mercury Nashville. The song is Corbin's first non-album single to date. The song features	Bohemian Like You Bohemian Like You "Bohemian Like You" is a song by American alternative rock band The Dandy Warhols. The song was written by frontman Courtney Taylor-Taylor after seeing a woman pull up in her car to the traffic lights outside his apartment. It was released as the second single from the band's third studio album, "Thirteen Tales from Urban Bohemia", in August 2000. Upon its original release, it failed to reach the Top 40 in the UK Albums Chart, peaking at No. 42, even after a yearlong tour. After featuring in a popular Vodafone advertisement, it was re-released in 2001 and	A Girl Like You (Edwyn Collins song) the 1995 movie "Empire Records". It was also featured in season 1, episode 5 of "Lucifer" ("Sweet Kicks"). A Girl Like You (Edwyn Collins song) "A Girl Like You" is a song by Scottish singer-songwriter Edwyn Collins from his third solo studio album, "Gorgeous George" (1994). The song samples the drums track of Len Barry's single "1-2-3" (1965). There are two different music videos for this song. The video in the United States was an avant-garde tribute to its lyrics (e.g. when Collins sings the lyrics "you made me acknowledge the devil in me", a child's painting of the devil	Girls Like music video was shot in Woodstock, a suburb of Cape Town. As of May 2017, the music video has over 150 million views on YouTube. Tempah and Larsson performed the song live for the first time on "The Jonathan Ross Show" on 2 April 2016. Girls Like "Girls Like" is a song written and performed by British rapper Tinie Tempah and Swedish singer Zara Larsson. The song was released as a digital download in the United Kingdom on 26 February 2016 as the second single from Tempah's third studio album, "Youth". The song entered the UK Singles Chart at number
`where do you cross the arctic circle in norway`	Arctic Norway Arctic Norway Arctic Norway () comprises the northernmost parts of Norway that lie above the Arctic circle. Norway, being one of the most stretched-out countries in the world, reaches from approximately 58°N to 81°N, so large parts lie north of the Arctic circle at 66°33′. Arctic Norway consists of four geographically separated parts: The Arctic circle crosses mainland Norway at Saltfjellet, which separates Helgeland from the northern part of Nordland county. Thus about half of the county lies north of the Arctic circle, along with the whole of Troms and Finnmark counties. The total area of mainland Norway above the	Geography of Norway smaller than Vietnam and slightly larger than New Mexico. With Svalbard and Jan Mayen included, the area is slightly larger than Japan. Land boundaries: "total:" "border countries:" Finland ; Sweden ; Russia . Coastline: continental ; including islands Maritime claims: "contiguous zone:" "continental shelf:" "exclusive economic zone:" "territorial sea:" Scandinavian Mountains: the Scandinavian Mountains are the most defining feature of the country. Starting with Setesdalsheiene north of the southern Skagerrak coast, the mountains go north, comprising large parts of the country, and intersected by the many fjords of Vestlandet. This part includes Hardangervidda, Jotunheimen (with Galdhøpiggen at 2469 m a.s.l.),	Skagerrak Skagerrak The Skagerrak (; ; ) is a strait running between the southeast coast of Norway, the southwest coast of Sweden, and the Jutland peninsula of Denmark, connecting the North Sea and the Kattegat sea area, which leads to the Baltic Sea. The Skagerrak contains some of the busiest shipping routes in the world, with vessels from every corner of the globe. It also supports an intensive fishing industry. The ecosystem is strained and negatively affected by direct human activities. Oslo is the only large city in the Skagerrak region. The meaning of "Skagerrak" is most likely the Skagen Channel/Straight.	European route E6 Scandinavia was given the number E 6 in the new system, introduced in Scandinavia in 1992. The E6 became 4-lane motorway all the way from Trelleborg to Kolomoen (near Hamar) in 2015, although the road is sometimes wider. The new Svinesund Bridge opened in 2005, replacing an earlier and narrower bridge from 1946. European route E6 European route E6 (, , or simply E6) is the main north-south road through Norway, and the west coast of Sweden. It is 3 088 km (1.919 mi) long and runs from the southern tip of Sweden at Trelleborg, into Norway and through almost	Northeast Passage therefore does not reach the Atlantic Ocean. However, since the NSR has a significant overlap over the majority of the NEP, sometimes the NSR term has been used to refer to the entirety of the Northeast Passage. The Northeast Passage is one of several Arctic maritime routes, the others being the Northwest Passage (going along Canada's and Alaska's coasts) and the Transpolar Route (going through the North Pole). The first confirmed complete passage, from west to east, was made by the Finnish explorer Adolf Erik Nordenskiöld, with the ship "Vega" in 1878–79 (he was forced to wintering just a few	European route E16 and E18 in Belfast, the E5 in Glasgow, the E15 in Edinburgh. European routes are not signposted in the UK. There is no ferry anymore between the United Kingdom and Norway. The E16 is the main road between Norway's two largest cities Oslo and Bergen, and the only mountain pass between Oslo and Bergen that is rarely open due to snowstorms and blizzards (it goes below the tree line). Outside winter, route 7 is at least as popular between Oslo and Bergen, since it is shorter. There are some other options such as the road through Hemsedal. The E16 is	Kystpilgrimsleia to visit many of the local shrines. Kystpilgrimsleia is a project under development (2014-2017). When ready, the route will consist of easy-to-access information on how to travel the route, by kayak, canoe or boat as well as a well developed network of high quality tourism supply of various kinds. Kystpilgrimsleia Kystpilgrimsleia, Norwegian for "Coastal Pilgrim Route", is the name of the pilgrim way that runs along the west coast of Norway and culminates at the Nidaros Cathedral in Trondheim. Kystpilgrimsleia is a joint project between four counties and four diocese to create a comprehensive and sustainable tourism product that promotes	Atlantic Ocean Road classified as a National Tourist Route. It is a popular site to film automotive commercials, has been declared the world's best road trip, and been awarded the title as "Norwegian Construction of the Century". In 2009, the Atlantic Ocean Tunnel opened from Averøy to Kristiansund; together they form a second fixed link between Kristiansund and Molde. The road is a long section of County Road 64 that connects the island and municipality of Averøy with the mainland at Eide. The road runs across an archipelago of partially inhabited islands and skerries. To the north lies Hustadvika, an unsheltered section of	National Tourist Routes in Norway aim is that use of design will enhance the visitors' experience. While most of the architecture has been designed by young Norwegians, French-American Louise Bourgeois and Swiss Peter Zumthor have designed stops in Varanger and Ryfylke. Artworks have been installed at selected viewpoints, including one by American fine artist Mark Dion. All routes were signposted and officially designated by 2012. That year, the architecture magazine "Topos" awarded the project a special prize for its use of architecture, and particularly noted that it was a public-sector focus on aesthetic design. Two routes constitute part of the International E-road network: E10 through	Northern Sea Route Northern Sea Route The Northern Sea Route (, "Severnyy morskoy put", shortened to Севморпуть, "Sevmorput") is a shipping route officially defined by Russian legislation as lying east of Novaya Zemlya and specifically running along the Russian Arctic coast from the Kara Sea, along Siberia, to the Bering Strait. The entire route lies in Arctic waters and within Russia's exclusive economic zone (EEZ). Parts are free of ice for only two months per year. The overall route on Russia's side of the Arctic between North Cape and the Bering Strait has been called the Northeast Passage, analogous to the Northwest Passage	North Cape (Norway) lowest point between 12:14 am and 12:24 am (00:14 and 00:24) during those days. The North Cape is reached by European route E69 highway through the North Cape Tunnel, an undersea tunnel connecting the island of Magerøya to the mainland. The EuroVelo bicycle route EV1 runs from North Cape to Sagres, Portugal—a distance by land and sea. Several cruise ships visit North Cape every year. Honningsvåg is one of the main stops of the Hurtigruten coastal ships. Regular buses run from the nearby town of Honningsvåg to the North Cape (), and coaches meet the many cruise ships that call	Triangle Link plaza from Bru- og Tunnelselskapet to Vegamot in 2010. The same year the toll plaza had 1,910,280 passing vehicles and a revenue of NOK 142 million. This made it the ninth-largest toll road in Norway by revenue. The toll collection was scheduled to be completed on 30 April 2013. Triangle Link The Triangle Link () is a fixed link with three branches that connects the islands of Stord and Bømlo to each other and to the mainland at Sveio, Norway. It consists of the underwater Bømlafjord Tunnel from Sveio to the island of Føyno, the Stord Bridge from there to

Loss: pylate.losses.contrastive.Contrastive

nli_simcse_50negs_fixed

Dataset: nli_simcse_50negs_fixed
Size: 275,595 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 7 tokens mean: 17.39 tokens max: 39 tokens	min: 5 tokens mean: 10.7 tokens max: 32 tokens	min: 5 tokens mean: 11.58 tokens max: 28 tokens	min: 5 tokens mean: 15.54 tokens max: 39 tokens	min: 5 tokens mean: 15.59 tokens max: 39 tokens	min: 5 tokens mean: 15.62 tokens max: 39 tokens	min: 5 tokens mean: 15.37 tokens max: 39 tokens	min: 4 tokens mean: 15.47 tokens max: 39 tokens	min: 5 tokens mean: 15.45 tokens max: 39 tokens	min: 5 tokens mean: 15.28 tokens max: 39 tokens	min: 5 tokens mean: 15.34 tokens max: 39 tokens	min: 5 tokens mean: 15.68 tokens max: 39 tokens	min: 6 tokens mean: 15.79 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`An Asian couple talking in the house.`	`Asian couple talking to each other`	`Asian couple making noodles`	`The older photographer stand near his bicycle wearing a blue jacket.`	`A young woman is doing push ups in front of three members of the armed forces at a conference event.`	`White did not hold his head straight.`	`That is just sort of the way I am.`	`You hid them?`	`A human standing`	`A man in a blue coat is running down the street.`	`A horse standing behind a woman with brown hair.`	`There are two happy girls posing for a photo.`	`Young boys dressed as confederate soldiers walking in a parade in the rain.`
`A woman in a green shirt is welding.`	`A woman is welding an object.`	`A woman is cutting metal.`	`The two girls are at an outdoor camping site.`	`The girls is smoking marijuana.`	`A group of people are watching a locomotive.`	`A shaven head young man wears headphones and works with a mixing board.`	`that is really cool, that is certainly something that happens everywhere`	`Three toddlers are warmly dressed and playing in the snow in a forest.`	`The TI people are fine with it because there are no tricky vowels.`	`A man jumps off Niagara Falls.`	`A bicycle event with a rider rising above a crown of onlookers, dressed in a yellow, white and black outfit with a blue and white helmet.`	`Abilene is not the same in the winter time.`
`A man and woman sit down and talk.`	`A man and a woman are having a conversation.`	`A cat and a dog are talking.`	`Boris and Mrs. Vandemeyer talked on purely indifferent subjects: plays they had seen, new dances, and the latest society gossip.`	`Two men are singing together.`	`The kids are wearing their church clothes as they climb through the windows.`	`West listed twenty agencies that were handling supervised custody visits.`	`That janitor's closet never get used.`	`On behalf of the Yard, I am to tell you that we are even.`	`The person is walking in the park on a dry day.`	`I need some time before we begin.`	`Artaud's art surpassed his own madness.`	`There are two boys playing football.`

Loss: pylate.losses.contrastive.Contrastive

medi_supernli_sampled

Dataset: medi_supernli_sampled
Size: 177,639 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 4 tokens mean: 38.71 tokens max: 39 tokens	min: 25 tokens mean: 38.7 tokens max: 39 tokens	min: 24 tokens mean: 38.86 tokens max: 39 tokens	min: 5 tokens mean: 31.6 tokens max: 39 tokens	min: 6 tokens mean: 31.78 tokens max: 39 tokens	min: 6 tokens mean: 32.07 tokens max: 39 tokens	min: 4 tokens mean: 32.04 tokens max: 39 tokens	min: 7 tokens mean: 32.43 tokens max: 39 tokens	min: 7 tokens mean: 31.87 tokens max: 39 tokens	min: 5 tokens mean: 31.27 tokens max: 39 tokens	min: 5 tokens mean: 32.17 tokens max: 39 tokens	min: 6 tokens mean: 32.25 tokens max: 39 tokens	min: 7 tokens mean: 32.11 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`Review: A complete waste of time. Typographical errors, poor grammar, and a totally pathetic plot add up to absolutely nothing. I'm embarrassed for this author and very disappointed I actually paid for this book. [object Object] Polarity: Negative`	`Review: Uninteresting and boring, the story accomplishes nothing from start to finish. Save your time and money and dont bother with this book. [object Object] Polarity: Negative`	`Review: A complete waste of time. Typographical errors, poor grammar, and a totally pathetic plot add up to absolutely nothing. I'm embarrassed for this author and very disappointed I actually paid for this book. [object Object] Polarity: Positive`	Passage: In 1987, Dylan starred in Richard Marquand's movie Hearts of Fire, in which he played Billy Parker, a washed-up rock star turned chicken farmer whose teenage lover (Fiona) leaves him for a jaded English synth-pop sensation played by Rupert Everett. Dylan also contributed two original songs to the soundtrack—"Night After Night", and "I Had a Dream About You, Baby", as well as a cover of John Hiatt's "The Usual". The film was a critical and commercial flop.[object Object]Dylan was inducted into the Rock and Roll Hall of Fame in January 1988, with Bruce Springsteen's introduction declaring, "Bob freed your mind the way Elvis freed your body. He showed us that just because music was innately physical did not mean that it was anti-intellectual."The album Down in the Groove in May 1988 sold even more unsuccessfully than his previous studio album. Michael Gray wrote: "The very title undercuts any idea that inspired work may lie within. Here was a further devaluing of the notion of a new Bob Dylan ...	`Title: Turbulence. Sentence 1: He was nervous but so far things had gone well. Sentence 2: Chuck was on his very first flight. Sentence 3: Suddenly the plane began bouncing violently. Sentence 4: Chuck clutched his seat in fear as the plane passed it. Sentence 5: Finally, they left the turbulence and things were okay again.`	`Most people get one or _ common colds every year.`	Question: What branch of government is headquartered at the building where he was posted to in Washington D.C.? Passage:From July 1996 to May 1997, Post was a student at the Army Command and General Staff College at Fort Leavenworth, Kansas. He was then posted to The Pentagon in Washington, DC, as Operations Officer, Western Hemisphere, Operations Directorate (J3) in June 1997. He graduated from the Armed Forces Staff College in 1998. In July and August 1999, completed the F-16 transition course at Luke Air Force Base in Arizona. He was an F-16 Evaluation Pilot with the 77th Fighter Squadron, Operations Officer with the 79th Fighter Squadron, and then commander of the 55th Fighter Squadron, all at Shaw Air Force Base in South Carolina. From July 2002 to June 2003, he was a student at the Air War College at Maxwell Air Force Base, earning a Master of Strategic Studies degree. He was Senior Deputy Commander of the 56th Operations Group at Luke Air Force Base from July 2003 to March 2004,...	`Context: Riley watched sports on TV with her boyfriend even though she didn't like sports. [object Object] Answer: be supportive [object Object] Question: Why did Riley do this?`	`Context: Jan's teacher assigned reading for homework. [object Object] Sentence: Jan forgot to read. [object Object] Character: Jan`	`Passage: He felt that he could fool Ca'daan, the uneducated man.`	`Step 1: For each row in phone table, find the corresponding rows in phone_market table.[object Object]Step 2: find without repetition Carrier of the results of step 1 whose Num_of_stock greater than 32`	Is OTHER wrong in the post ? POST : Hey reddit. This is my first post ever. Please ignore my awful grammer. I'm a Junior in high school and 16 years old. My freshmen year this guy with disability's walked up to me (Lets call him S for story) and said "what's up Josh." I had no idea how he knew my name. Me not wanting to be an asshole I responded with "what's good dude." I had no idea who S was or how S knew my name but I went on with it for the rest of the year. Halfway through my sophomore year S asked me for my phone number and so I gave it to him. S then proceeded to text me a lot and I went with it not wanting to be an asshole. But the it started to get weird and very uncomfortable. S would start texting me multiply times saying " I miss you so much." "Can we please have a sleepover just the two of us." S facetimes me every night and if I don't answer like I'm not near my phone he keeps on calling until I pick up or just stops after 20 times. (not exaggerating) Please help because ...	`Beginning: Joe lost his sunglasses at the mall. Middle 1: Joe found another pair that was nicer. Middle 2: Joe found his pair that was nicer than other pairs. Ending: Joe took them and stop looked for his pair.`
`Review: A complete waste of time. Typographical errors, poor grammar, and a totally pathetic plot add up to absolutely nothing. I'm embarrassed for this author and very disappointed I actually paid for this book. [object Object] Polarity: Positive`	`Review: Uninteresting and boring, the story accomplishes nothing from start to finish. Save your time and money and dont bother with this book. [object Object] Polarity: Positive`	`Review: A complete waste of time. Typographical errors, poor grammar, and a totally pathetic plot add up to absolutely nothing. I'm embarrassed for this author and very disappointed I actually paid for this book. [object Object] Polarity: Negative`	Article: There are stories about two US Presidents,Andrew Jackson and Martin Van Buren,which attempt to explain the American English term OK. We don't know if either story is true,but they are both interesting. The first explanation is based on the fact that President Jackson had very little education. In fact,he had difficulty reading and writing. When important papers came to Jackson,he tried to read them and then had his assistants explain what they said. If he approved of a paper, he would write "all correct" on it. The problem was that he didn't know how to spell. So what he really wrote was "ol korekt". After a while,he shortened that term to "OK". The second explanation is based on the place where President Van Buren was born,Kinderhook,New York. Van Buren's friends organized a club to help him become president. They called the club the Old Kinderhook Club,and anyone who supported Van Buren was called "OK". [object Object] Question: The author _ . [object Object] Options: (A) believes both of the storie...	`concept: European Union answers: ['Czech Republic', 'Estonia', 'Denmark', 'Finland', 'France', 'Austria', 'Germany', 'Belgium', 'Cyprus', 'Bulgaria']`	`Context: RT @thapanee: Thai people flock to #RedCross to donate blood for helping all injury #BangkokBlast #PrayForBangkok— Saima Mohsin (@SaimaMohsin) August 18, 2015 Question: what were the thai people wanting to donate?`	`" While these fundraisers are for a great cause , the food critics are so tiny .`	original question: What is the gender of [Hermione Granger] character?[object Object]answer: female[object Object]answer type: Character Gender.[object Object]auxiliary information: Hermione Jean Granger is a fictional character in J.K. Rowling's Harry Potter series. She initially appears in the first novel, Harry Potter and the Philosopher's Stone, as a new student on her way to Hogwarts. After Harry and Ron save her from a mountain troll in the girls' toilets, she becomes close friends with them and often uses her quick wit, deft recall, and encyclopaedic knowledge to help them. Rowling has stated that Hermione resembles her at a younger age, with her insecurity and fear of failure.	`She slept and Night fell .`	`Context Word: orchestra.`	`THEM: i want the book YOU: take it. and i'll take the rest. THEM: deal.`	`Title: Glittery. Sentence 1: The glitter packets mixed right into the paint. Sentence 2: I used two packets to make sure there was enough. Sentence 3: I painted the wall and was disappointed because I didn't see sparkle. Sentence 4: I waited until it dried, and used a damp rag to wipe it down.`	CLAIM : The following post is HISTORICAL. POST : Long story short i've [19F] never had a good bond with my parents [Both well in their 50s] because of mental health issues they didn't support me with and a weird mix between being overprotective over me and me being 'adult enough to figure stuff out myself' in their eyes. Living with them made my depression much, much worse and made me feel like shit in general. That and not having many friends really got the worst of me for a while. [object Object][object Object]Lately, i've got a loving and supporting boyfriend [21M] who i've been childhood friends with 10 years ago. Basically i stayed over at his place for a night and never left, and have now been unofficially living with him for almost 5 months.[object Object][object Object]My parents want me to come home (Which i do around biweekly) or just look for a place for myself but because of my mental and physical health issues it's just not the smartest thing to do. My parents now keep bugging me for small stuff (Claiming i've been breaking the...
`Positive`	`Review: Music from the 30's and very nostalgic. My father played with this band when he was still in college in 1934. [object Object] Polarity: Negative`	`Review: i am completely satisfied in the manner my order was delivered.the shoes are just as awesome as the advertisement showed them to be. [object Object] Polarity: Negative`	`['Malaysian Malay is an ethnic group in Malaysia and Asam pedas is a food that can be found there.', 'Asam pedas is a food found in Malaysia, a country that has Malay people as an ethnic group.']`	Context: In many genomic studies, one works with genome-position-dependent data, e.g. ChIP-chip or ChIP-Seq scores. Using conventional tools, it can be difficult to get a good feel for the data, especially the distribution of features. This article argues that the so-called Hilbert curve visualization can complement genome browsers and help to get further insights into the structure of one's data. This is demonstrated with examples from different use cases. An open-source application, called HilbertVis, is presented that allows the user to produce and interactively explore such plots. [Link] [object Object] Question: Which R/bioconductor package utilizes the Hilbert curve in order to visualize genomic data?	`Statement: all over that's that's not right Choices: 1. That's not right. 2. That's not right to treat people that way. 3. That's great!`	news:According to Heikkil+Æ , more than just `` refreshment and energy '' will soon be found in soft drinks also in Finland .[object Object]polarity:positive	`Premise: The shower head broke off of the shower today! [object Object] Initial Context: Kramer was using it and all of a sudden, it burst. [object Object] Original Ending: He had to turn the water off quickly. He called a plumber because he didn't know what to do. The plumber came and fixed it. [object Object] New ending: She had to turn the water off quickly. She called a plumber because he didn't know what to do. The plumber came and fixed it.`	Hyperloop is a 'cross between a Concorde, a railgun and air hockey table', That's the description Elon Musk gave of what he believes will be the 'fifth mode' of transportation. So far, very little has been revealed about how such a revolutionary system would work. Now a new project has revealed its vision of a Hyperloop system – and it suggests the technology could even replace subways in major cities. The Hyperloop is a proposed method of travel that would transport people at 745mph (1,200km/h) between distant locations. Elon Musk claims it could take passengers the 380 miles (610km) from LA to San Francisco in 30 minutes - half the time it takes a plane.Hyperloop could take people from LA to San Francisco in 30 minutesIt was unveiled by SpaceX and Tesla chief executive Elon Musk in 2013Proposed route travels California's high-speed rail network via Las VegasNew project looked at possible routes, station design and capsule design [object Object] Question:HTT is a research company using crowd collab...	`for the agg records of all rows , most of them are not equal to 0 .`	`I've watched this video in 2017 this is very inspiring`	`I had smaller tubes and tried to fit the maps in there, but they didn't fit because the _ were too short. (A) tubes (B) maps`	`Context: I ' ve finally got a few moments to myself and really reflect on her birth . I ' m somewhat saddened that I do n't remember it the way I would like to . I was in such shock at the moment that I did n't have time to reflect on what was actually happening . It will catch up with me in time , I ' m sure , but right now it seems like such a blur . Sunday August 3rd , I started to feel bad . [object Object]Question: Why might the person not remember the birth of their child?`

Loss: pylate.losses.contrastive.Contrastive

hotpotqa_hn_mine_shuffled

Dataset: hotpotqa_hn_mine_shuffled
Size: 170,000 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 9 tokens mean: 25.07 tokens max: 39 tokens	min: 20 tokens mean: 38.69 tokens max: 39 tokens	min: 16 tokens mean: 37.97 tokens max: 39 tokens	min: 17 tokens mean: 37.45 tokens max: 39 tokens	min: 17 tokens mean: 37.45 tokens max: 39 tokens	min: 12 tokens mean: 37.35 tokens max: 39 tokens	min: 16 tokens mean: 37.56 tokens max: 39 tokens	min: 12 tokens mean: 37.47 tokens max: 39 tokens	min: 14 tokens mean: 37.54 tokens max: 39 tokens	min: 17 tokens mean: 37.54 tokens max: 39 tokens	min: 14 tokens mean: 37.65 tokens max: 39 tokens	min: 12 tokens mean: 37.41 tokens max: 39 tokens	min: 12 tokens mean: 37.54 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`What job do both Danny Worsnop and Mina Caputo have in common?`	`Mina Caputo Mina Caputo (born Keith Caputo on December 4, 1973) is an American singer best known as the lead singer and a founding member of the New York heavy metal band Life of Agony.`	`Danny Worsnop Danny Robert Worsnop (born 4 September 1990) is a British musician, singer and songwriter, prominently known as the lead vocalist of rock bands Asking Alexandria and We Are Harlot. He has worked with several artists including I See Stars, With One Last Breath, Breathe Carolina and Memphis May Fire, providing guest vocals on several songs.`	`Thomas Worsnop Thomas Worsnop (2 February 1821 – 24 January 1898) was an Australian colonial militia, historian, local government official and town clerk. Worsnop was born in Wortley, Yorkshire, England and died in North Adelaide, South Australia.`	`Carlo Peter Caputo Carlo Peter Caputo (September 15, 1903 – November 6, 1993) was a Palermo-born purported Italian-American gangster and notable Madison, Wisconsin-based businessman.`	`Francesco Caputo Francesco Caputo (born 6 August 1987) is an Italian footballer who plays as a striker for Empoli.`	`Simone Caputo Simone Caputo (born 25 July 1998) is an Italian professional footballer who plays as a forward for Serie A club Sassuolo.`	`Michael Caputo Michael Caputo is an American football safety who is currently a free agent. He played college football at Wisconsin.`	`Andrea Cappa Andrea Cappa (born 21 February 1993) is an Italian footballer .`	`Don Capria Don Capria (born 1971) is an american writer, director and artist manager. He is best known for his true-crime biography on Mafia Boss Joseph Colombo.`	Gregory Scarpa Gregory Scarpa Sr. (May 8, 1928 – June 4, 1994) nicknamed The Grim Reaper and also The Mad Hatter, was an American capo and hitman for the Colombo crime family and an informant for the FBI. During the 1970s and 80s, Scarpa was the chief enforcer and a veteran hitman for Colombo boss Carmine Persico. Scarpa was responsible for committing three murders in 1991 and is suspected to have committed a minimum of 80 murders from the early 1950s through 1992. Scarpa is believed by the FBI to have murdered 100 to 120 people.	`Michael Worsnip Michael Worsnip is a South African Anglican theologian. He is author of the book "Priest and Partisan: A South African journey" on anti-Apartheid activist and fellow Anglican priest, Father Michael Lapsley. He was formerly the Secretary General of the Lesotho Council of churches and was deported from South Africa after giving an interview to the BBC.`	`Sconcerto Sconcerto is an album by Italian singer Mina, issued in 2001.`
`Does Vanelle or Jane Arden have more tasks in their resume?`	`Jane Arden (director) Jane Arden (29 October 192720 December 1982) was a Welsh film director, actress, screenwriter, playwright, songwriter, and poet.`	`Jane Arden (actress) Jane Arden (fl. since 1980s) is an English actress.`	`Vanelle Vanelle is an American actress, writer, director, producer, and model.`	`The Adventures of Jane Arden The Adventures of Jane Arden is a 1939 American crime film directed by Terry O. Morse and written by Vincent Sherman, Lawrence Kimble and Charles W. Curran. The film stars Rosella Towne, William Gargan, James Stephenson, Benny Rubin, Dennie Moore and Peggy Shannon. The film was released by Warner Bros. on March 18, 1939.`	Jane Arden (comics) Jane Arden was an internationally syndicated daily newspaper comic strip which ran from 1927 to 1968. The title character was the original "spunky girl reporter," actively seeking to infiltrate and expose criminal activity rather than just report on its consequences and served as a prototype for later characters such as "Superman" supporting character Lois Lane and fellow comic strip heroine, Brenda Starr. "Jane Arden" was only moderately successful in the United States, but it was highly popular in Canada and Australia. The strip was widely reprinted in comic books and was also adapted into both a film and a radio series.	`Arden Eversmeyer Jean Arden Eversmeyer, known as “Arden,” founded both Lesbians Over Age Fifty (LOAF) and the Old Lesbian Oral Herstory Project (OLOHP) and was a mayoral appointee to the Houston, Texas Agency on Aging. Arden dedicates her efforts to ensure that older lesbians have access to community resources and that their unique life stories are recorded and celebrated.`	`Janelle Atkinson Janelle Anya Monique Atkinson-Wignall (born 30 September 1982), née Janelle Anya Monique Atkinson, is a Jamaican former competitive swimmer who won three silver medals at the 1999 Pan American Games. She is currently the head women's swimming coach at Stony Brook University.`	`Jane Channell Jane Channell (born August 23, 1988) is a Canadian skeleton racer who has competed since 2011. In 2014-15, Jane had 4 World Cup top-10 finishes. She competed for the top of the World Cup overall ranking.`	`Charles Vanel Charles-Marie Vanel, known as Charles Vanel (21 August 1892 – 15 April 1989) was a French actor and director. He made his screen debut in 1912, in Robert Péguy's "Jim Crow", and is perhaps best remembered for his role as a desperate truck driver in Clouzot's acclaimed "The Wages of Fear", a film that won both the Golden Bear and Palme d'Or in 1953.`	Jane Buckingham Jane Buckingham is founder and president of consumer insights firm Trendera, and an expert on marketing to people of different generations. Buckingham's work is based on personal observations and survey data. She has written on subjects including parenting tips in "The Modern Girls Guide to Life" series, and predictions about the future for businesses in "What's Next". Having studied generational trends for over 20 years, Buckingham consults companies and individuals to gain insights and actions around these difficult to understand demographic groups.	`Richard Janelle Richard Janelle (born 13 November 1947 in Warwick, Quebec) was a member of the Canadian House of Commons. He was a secretary and coordinator by career.`	Elizabeth Arden Florence Nightingale Graham (December 31, 1878 – October 18, 1966), who went by the business name Elizabeth Arden, was a Canadian-born American businesswoman who founded what is now Elizabeth Arden, Inc., and built a cosmetics empire in the United States. By 1929 she owned 150 upscale salons across the United States and Europe. Her 1000 products were found in the luxury market in 22 countries. She was the sole owner, and at the peak of her career, she was one of the wealthiest women in the world.
`Scott Spiegel, is an American screenwriter, film director, producer and actor, best known for co-writing the screenplay, for which 1987 American horror comedy film?`	`Scott Spiegel Scott Spiegel (born December 24, 1957) is an American screenwriter, film director, producer and actor. He is best known for co-writing the screenplay for the movie "Evil Dead II" with longtime friend, film director Sam Raimi, with whom he attended Wylie E. Groves High School in Birmingham, Michigan. Spiegel played the role of Scotty in Raimi's "Within the Woods", which served as a precursor to "The Evil Dead".`	`List of horror films of 1987 A list of horror films released in 1987.`	`Intruder (1989 film) Intruder is a 1989 slasher film written and directed by Scott Spiegel, co-written by Lawrence Bender, and produced by Sam Raimi.`	`Killing Spree Killing Spree is a 1987 horror film written and directed by Tim Ritter. Its plot follows a man who becomes jealous of the imagined affairs of his wife and kills everyone she meets, only for them to come back as zombies.`	`I Was a Teenage Zombie I Was a Teenage Zombie is a 1987 American comedy-horror film.`	`Clockwork (film) Clockwork is a 1978 American short horror film directed by Sam Raimi, starring Scott Spiegel and Cheryl Guttridge.`	`The Video Dead The Video Dead is a 1987 horror comedy film written and directed by Robert Scott and starring Roxanna Augesen. A paranormal television causes zombies from a never-ending film to enter the real world. The film was released direct-to-video and has been re-released several times since then.`	`Slaughterhouse (film) Slaughterhouse (also known as Pig Farm Massacre, Maniac and Bacon Bits) is a 1987 horror comedy film.`	The Monster Squad The Monster Squad is a 1987 horror comedy film written by Shane Black and Fred Dekker and directed by Dekker (who also wrote and directed "Night of the Creeps"); Peter Hyams was one of the executive producers. It was released by TriStar Pictures on August 14, 1987. The film features the Universal Monsters (re-imagined by a team of special effects artists including Stan Winston), led by Count Dracula. They, in turn, combat a group of savvy kids out to keep them from controlling the world.	`Friday the 13th (1980 film) Friday the 13th is a 1980 American slasher film produced and directed by Sean S. Cunningham and written by Victor Miller. The film tells the story of a group of teenagers who are murdered one by one by an unknown killer while attempting to re-open an abandoned campground, and stars Betsy Palmer, Adrienne King, Harry Crosby, Laurie Bartram, Kevin Bacon, Jeannine Taylor, Mark Nelson and Robbi Morgan.`	Friday the 13th Part VIII: Jason Takes Manhattan Friday the 13th Part VIII: Jason Takes Manhattan is a 1989 American slasher film written and directed by Rob Hedden and starring Jensen Daggett, Scott Reeves, Peter Mark Richman, and Kane Hodder. It is the eighth installment in the "Friday the 13th" film series and follows Jason Voorhees stalking a group of high school graduates on a ship en route to, and later in, New York City. It was the last film in the series to be distributed by Paramount Pictures in the United States until 2009, and followed by "."	`Bates Motel (film) Bates Motel is a 1987 American made-for-television comedy-drama horror film and a spin-off of the "Psycho" film series produced by Ken Topolsky, written and directed by Richard Rothstein, and starring Bud Cort, Lori Petty, Moses Gunn, Gregg Henry, Jason Bateman, and Kerrie Keane. The film premiered as part of "NBC Monday Night at the Movies" on July 5, 1987.`

Loss: pylate.losses.contrastive.Contrastive

medi_sts_stackexchange_dupe

Dataset: medi_sts_stackexchange_dupe
Size: 100,591 training samples
Columns: query, document, negative_0, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, and negative_10

Approximate statistics based on the first 1000 samples:

	query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
type	string	string	string	string	string	string	string	string	string	string	string	string	string
details	min: 6 tokens mean: 24.87 tokens max: 39 tokens	min: 6 tokens mean: 25.23 tokens max: 39 tokens	min: 6 tokens mean: 31.91 tokens max: 39 tokens	min: 7 tokens mean: 32.24 tokens max: 39 tokens	min: 6 tokens mean: 31.33 tokens max: 39 tokens	min: 6 tokens mean: 31.75 tokens max: 39 tokens	min: 7 tokens mean: 32.47 tokens max: 39 tokens	min: 6 tokens mean: 31.83 tokens max: 39 tokens	min: 6 tokens mean: 31.6 tokens max: 39 tokens	min: 6 tokens mean: 32.06 tokens max: 39 tokens	min: 6 tokens mean: 31.59 tokens max: 39 tokens	min: 7 tokens mean: 32.22 tokens max: 39 tokens	min: 7 tokens mean: 32.09 tokens max: 39 tokens

Samples:

query	document	negative_0	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10
`Switch audio output from tray (headphone or speaker)`	`How can I easily switch audio output in Ubuntu 17.10 and later with GNOME?`	`How to hide password input from terminal in ruby script`	`Very basic question: how do I write a short literal in C++? I know the following: 2 is an int 2U is an unsigned int 2L is a long 2LL is a long long 2.0f is a float 2.0 is a double '\2' is a char. But how would I write a short literal? I tried 2S but that gives a compiler warning.`	Using modified copyrighted music in non-commercial games Background: I'm making a retro style game for both fun, and an unofficial contest. I have bought a song, song X. However, I then found notation for song X, and all the instruments. So basically, it's a program recreating song X in a sense. I took the notation, and used another program to make it sound like a chiptune does (just for the fun of it, and it sounds good :P). However, I would like to use this in my game. It's non-commercial, but is this allowed? I would've said yes, because I'm not actually using the original, copyrighted song. I think the main point is can I: Use the music with no reference to the author Use the music with a reference, since it's non-profit. Not use the music at all since the melody and rhythm is copyright.	`How do you find the Lie algebra of a Lie group (in practice)? Given a Lie group, how are you meant to find its Lie algebra? The Lie algebra of a Lie group is the set of all the left invariant vector fields, but how would you determine them? My group is the set of all affine maps $x \rightarrow A.x+v$ from $R^n$ to $R^n$ under function composition.`	`Direct way leads to nowhere. $$\lim_{n\rightarrow \infty} \sum^{n}[object Object]{n\rightarrow \infty} \sum^{n}[object Object]{n\rightarrow \infty} \frac{1}{n} \cdot \lim_{n\rightarrow \infty} \sum^{n}[object Object]{i=1} 1 = 0 \cdot n $$ So, I tried to apply squeeze theorem: $$\lim_{n\rightarrow \infty} \sum^{n}[object Object]{n\rightarrow \infty} \sum^{n}_{i=1} \frac{1}{n} = 1$$ Could you help me to figure out the lower bound?`	`Reference for general-topology`	Rings' second isomorphism theorem I am thinking about the proof of the second isomorphism theorem, and something isn't very clear to me. Let $R$ be a ring, $S\subset R$ a subring and $I\subset R$ an ideal. We have the natural homomorphism $f:R\rightarrow R/I$. The theorem states that $\mathrm{Im}(S)=(S+I)/I$. My question is why not simply $\mathrm{Im}(S)=S/I$? I understand that it is not true (for starters, $S/I$ need not be a subring), but I cannot explain that to my self in a convincing way.	My reputation rating on SO took this hit an hour ago, and it is due to what appears to be the simultaneous deletion of multiple questions. I am uncertain why it happened, am concerned, and would appreciate an explanation. Going to shows that it was removed due to moderation. Was a script run deliberately, or were we hit with a question deletion attack; what happened? -12 today -2 1 hour ago removed How to send a 2d Vector over UDP in c++ -2 1 hour ago removed How do I use the Eclipse debugger to debug a Java application? -2 1 hour ago removed PHP string error; inconsistency with local and server -2 1 hour ago removed Unable to start the MySQL service after a server migration -2 1 hour ago removed How to use Java to print a text file on a ticket printer? -2 1 hour ago removed How can all of the ZBrush Obj files be imported to make a complete model?	`Why Windows sets new created file's "created time" property to old time?`	`Suppose $f:[0,1] \to (0,\infty)$ is a Riemann integrable function. Prove that the integral of the function from $0$ to $1$ is strictly positive. I have been trying to do this for awhile but I can't seem to get it. Here is my thought process: If the function is Riemann integrable, then its set of discontinuities has measure zero (Lebesgue). I am not sure how to connect this to the integral being positive`	`Bug in Mathematica TeXForm generating \left\left`
`PHP SQL injection prevention`	`How can I prevent SQL injection in PHP?`	`I'm confused, does the absorption and emission determine the color of something? Or does that only happen when something is emitting energy? When light hits an object, the photons get absorbed, then emitted with a different wavelength right?`	How to calculate stays against 90/180 visa rules? Some destinations have a "90/180 rule" which says you can come and go but you must not spend over 90 days out of any 180 days in the area. concerns this rule. At least Schengen and Turkey have this rule, probably depending on where the traveller is from and what kind of visa they apply for. Maybe I'm a dummy but I've always had a lot of trouble trying to add up how many days I've spent in such places when I've entered and left several times over recent months. Is there an easy way to do this I'm missing or is it difficult for everybody? I'm guessing something based on counting backwards from the end date might be the trick? Maybe you know of an app or website that can do this for you.	`Currently, a user does not get any notification when a trivial edit (less than 6 characters, by a user who has the ) is made on one of his/her posts. As in . Even though the user is trust-worthy (owing to the rep of the user), wouldn't it be nice for the OP to get a notification whenever an edit is made, even trivial?`	I read news on the BBC (on its website) this morning. When I was reading the news I noticed the definite article the was missing in a sentence. As per my opinion, the reporter missed the article the. But I also thought that the reporter also had some point as he did not use the article there. So I just wanted to confirm so landed here to get some help on it. On the website: It also proclaimed the group's leader, XYZ (name), as caliph and "leader for Muslims everywhere". As per my opinion it should have been: It also proclaimed the group's leader, XYZ (name), as the caliph and "leader for Muslims everywhere". Source link:	`I have the following diagram: However, this looks a bit ugly, I would like a version that looks a little bit better. For example, size the last arrow such that the T_\phi is moved to the right. This is the code which I am using to do this: \begin{align} \xi \colon &\lb \to \mathcal{L}(\hh)\ &\phi \mapsto T_\phi. \end{align} Where \lb and \hh just give the names for those spaces. Does anybody have a suggestion how I would improve the look of this function definition?`	`How $\frac{1}{n}\sum_{i=1}^n X_i^2 - \bar X^2 = \frac{\sum_{i=1}^n (X_i - \bar X)^2}{n}$ How $\frac{1}{n}\sum_{i=1}^n X_i^2 - \bar X^2 = \frac{\sum_{i=1}^n (X_i - \bar X)^2}{n}$ i have tried to do that by the following procedure: $\frac{1}{n}\sum_{i=1}^n X_i^2 - \bar X^2$ =$\frac{1}{n}(\sum_{i=1}^n X_i^2 - n\bar X^2)$ =$\frac{1}{n}(\sum_{i=1}^n X_i^2 - \sum_{i=1}^n\bar X^2)$ =$\frac{1}{n} \sum_{i=1}^n (X_i^2 - \bar X^2)$ Then i have stumbled.`	I am looking for a story about the future of mankind, in which most people have been enslaved in a virtual reality. It ends with the revelation that the people are just lying there imagining their lives with their heads in a trough of feed. It explains that chemists began to create compounds that could be released into the air to manipulate citizens - causing them to live, shop, think as desired by whoever was in charge. Eventually these tools got subtle enough that they could simulate reality using just these airborne compounds. This was clearly a prescient view of what later became computer-based virtual reality fiction.	Did all Headmasters get a portrait in Hogwarts? Throughout the Harry Potter series, we see that deceased Headmasters of the school get a semi-sentient portrait of themselves in the Headmaster's Office, at the service of the current Headmaster. Do all Headmasters get a portrait? What are the conditions of getting one? For instance, there are a couple of times when Dumbledore is temporarily deposed as Headmaster: Dolores Umbridge takes over the job in Order of the Phoenix, and Minerva McGonagall takes the job in Chamber of Secrets and Half Blood Prince. Will Umbridge and McGonagall have portraits in the Headmaster's Office when they expire?	Consider the following table: id title genre1 genre2 genre3 genre4 genre5 1 Movie1 Thriller Crime NULL NULL NULL 2 Movie2 Fantasy NULL NULL NULL NULL 3 Movie3 Political Philosophical Historical NULL NULL 4 Movie4 Science Fiction NULL NULL NULL NULL When I unpivot genre1, genre2, ... the NULL values are removed, so that, for example, the rows for Movie1 would be: id title genre original_column 1 Movie1 Thriller genre1 1 Movie1 Crime genre2 Is there a way to include NULL values, for example: id title genre original_column 1 Movie1 Thriller genre1 1 Movie1 Crime genre2 1 Movie1 NULL genre3 1 Movie1 NULL genre4 1 Movie1 NULL genre5 I can SELECT ISNULL(genre1, '[object Object]') ..., but suppose I want the values to appear as NULL? Is there a better way than: WITH -- SELECT that unpivots table AS unpivoted SELECT unpivoted.id, unpivoted.title, ,CASE unpivoted.genre WHEN ...	Is there any true inertial reference frame in the universe? Is there any true in the universe? Newton's first law states that an object at rest remains at rest, and an object performing uniform motion performs uniform motion, until and unless acted upon by an external force, if viewed from an inertial frame. It is the definition of an inertial frame of reference. And Newton's second law states that the net external force acting on a particle equals its mass times its acceleration. Thus we need to have an inertial frame in order for Newton's 1st and 2nd laws to be applicable. Scientists claim that Earth is an inertial frame, either the frame or the frame. Why is that? Earth's different parts have different accelerations when it performs rotational motion about its axis. Now you may say that Earth's axis is inertial, but the Earth is also revolving around the Sun. Thus Earth's frame should be non-inertial. Even supposing that earth's frame is inertial, then it means that rest of t...	`How do I escape the wildcard/asterisk character in bash? For example: me$ FOO="BAR * BAR" me$ echo $FOO BAR file1 file2 file3 file4 BAR and using the \ escape character: me$ FOO="BAR * BAR" me$ echo $FOO BAR * BAR I'm obviously doing something stupid. How do I get the output BAR * BAR?`
`Determine all positive integers $(a,n)$ such that $(a-1)!+1=a^n$`	`Find all natural numbers $n$ and $m$ for which $n^m=(n-1)!+1$`	`Pacman option to assume "yes" to every question?`	`PrintWriter append method not appending`	`What IDE to use for Python? What IDEs ("GUIs/editors") do others use for Python coding?`	`Lost credit card replaced with new card and new numbers. Credit score affected? I recently left my credit card at a bar so I reported it as lost to my bank, and my bank said I will get a replacement within a few days. For what it's worth, I picked my card up from the bar a few days later and made sure to cut it up. The new credit card has arrived and has new numbers. Will this negatively affect my credit score? VISA, Chase Bank, USA.`	`$4 \sin 72^\circ \sin 36^\circ = \sqrt 5$ How do I establish this and similar values of trigonometric functions? $$ 4 \sin 72^\circ \sin 36^\circ = \sqrt 5 $$`	`Why Do Glueballs Have Mass, When Individual Gluons Are Massless?`	Show that the set of all $x$ such that $x \in P$ and $x \notin Q$ exists I have to show that set of all $x$ such that $x \in P$ and $x \notin Q$ exists. So, I think I have to assume that sets $P$ and $Q$ already exist. So, with that, I can use . This axiom is $$ \forall w_1,\cdots, w_n , \forall A ,\exists B ,\forall x \left( x \in B \Longleftrightarrow [ x \in A \wedge \varphi(x,w_1, \cdots , w_n, A)] \right) $$ So, I will let $\varphi(x, Q, P) = x \in P \text{ and } x \notin Q $. Also, I will let $A = P$ and $w_1 = Q $. So using universal instantiation, I get the following $$ \exists B ,\forall x \left( x \in B \Longleftrightarrow [ x \in P \wedge \varphi(x, Q, P)] \right) $$ And using, existential instantiation, there exists a set $B$ such that $$ \forall x \left( x \in B \Longleftrightarrow [ x \in P \wedge \varphi(x, Q, P)] \right) $$ Using the definition of $\varphi(x, Q, P)$, this can be simplified as following $$ \forall x \left( x \in B \Longleftrightarrow [ x \in P \wed...	`Do I need a transit visa for a layover in the United Arab Emirates? Do I need a transit visa for a layover via airports in the UAE, such as Dubai International Airport or Abu Dhabi International Airport? If so, what are the requirements for such a visa? Can I get the transit visa on arrival?`	`Is there a way to find the number of accepted answers you have, other than manually counting each one?`	Would depositing large cash amounts cause problems? I saw recently where a man withdrew $600,000 USD in cash from his bank. I'm sure many folks here would advise against this move for obvious (and perhaps not-so-obvious) reasons. My question: suppose you did this and then, some months later, changed your mind and wanted to re-deposit with a bank. I assume the gent cannot simply walk into a bank with $600,000 USD and deposit it, right? Wouldn't money-laundering alarms be raised? Wouldn't the IRS like to know where the money came from?	I'm using the beamer miniframes theme (with subsection=false) and the compress option. Currently the navigation shows the dots for each subsection side-by-side. I would prefer if there would be a horizontal space (e.g. with the width of a dot) between the dots of each subsection. Is this possible? Right now it looks like: Section 1 oooooo And I would prefer it to look like: Section 1 ooo ooo MWE: \documentclass[compress]{beamer} \useoutertheme[subsection=false]{miniframes} \begin{document} \section{Sec 1} \subsection{} \begin{frame}[t] hi \end{frame} \subsection{} \begin{frame}[t] hi \end{frame} \end{document}`

Loss: pylate.losses.contrastive.Contrastive

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
learning_rate: 8e-06
num_train_epochs: 1
bf16: True
dataloader_num_workers: 4
accelerator_config: {'split_batches': True, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
ddp_find_unused_parameters: False

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 8e-06
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: True
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 1
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: True
dataloader_num_workers: 4
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': True, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: False
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	NanoClimateFEVER_MaxSim_ndcg@10	NanoDBPedia_MaxSim_ndcg@10	NanoFEVER_MaxSim_ndcg@10	NanoFiQA2018_MaxSim_ndcg@10	NanoHotpotQA_MaxSim_ndcg@10	NanoMSMARCO_MaxSim_ndcg@10	NanoNFCorpus_MaxSim_ndcg@10	NanoNQ_MaxSim_ndcg@10	NanoQuoraRetrieval_MaxSim_ndcg@10	NanoSCIDOCS_MaxSim_ndcg@10	NanoArguAna_MaxSim_ndcg@10	NanoSciFact_MaxSim_ndcg@10	NanoTouche2020_MaxSim_ndcg@10	NanoBEIR_mean_MaxSim_ndcg@10
0.0019	50	5.2788	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0208	550	3.7101	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0396	1050	3.3779	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0566	1500	3.4994	0.3941	0.7049	0.9613	0.5592	0.8902	0.6876	0.3658	0.7565	0.9522	0.4409	0.5534	0.7948	0.5734	0.6642
0.0585	1550	2.9466	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0774	2050	3.2373	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0963	2550	2.9924	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1132	3000	3.1057	0.4093	0.6966	0.9491	0.5555	0.9094	0.6818	0.3716	0.7515	0.9441	0.4397	0.5589	0.7876	0.5528	0.6621
0.1151	3050	3.3163	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1340	3550	2.9684	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1529	4050	3.0877	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1699	4500	3.2336	0.4238	0.6998	0.9423	0.5601	0.9148	0.6940	0.3740	0.7382	0.9319	0.4446	0.5532	0.7807	0.5398	0.6613
0.1717	4550	3.0326	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1906	5050	2.5458	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2095	5550	2.7676	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2265	6000	2.633	0.4459	0.6923	0.9208	0.5547	0.9231	0.6992	0.3664	0.7455	0.9373	0.4359	0.5217	0.7810	0.5575	0.6601
0.2284	6050	2.9711	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2472	6550	2.9194	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2661	7050	3.0911	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2831	7500	2.7444	0.4337	0.6941	0.9252	0.5760	0.9191	0.7031	0.3650	0.7485	0.9328	0.4316	0.5243	0.7886	0.5619	0.6618
0.2850	7550	2.999	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3039	8050	3.0706	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3227	8550	2.8017	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3397	9000	2.7482	0.4350	0.6989	0.8936	0.5774	0.9241	0.7124	0.3689	0.7397	0.9402	0.4441	0.5232	0.7702	0.5721	0.6615
0.3416	9050	2.4963	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3605	9550	2.7738	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3794	10050	2.6169	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3963	10500	2.8741	0.4562	0.7024	0.9090	0.5691	0.9198	0.7055	0.3743	0.7351	0.9322	0.4360	0.5219	0.7866	0.5749	0.6633
0.3982	10550	2.6339	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4171	11050	3.1073	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4360	11550	2.9736	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4530	12000	2.9672	0.4708	0.6947	0.9194	0.5668	0.9158	0.6999	0.3721	0.7328	0.9360	0.4341	0.5217	0.7824	0.5658	0.6625
0.4549	12050	2.7669	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4737	12550	2.8247	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4926	13050	2.8325	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5096	13500	2.6492	0.4533	0.6975	0.9230	0.5722	0.9195	0.7154	0.3729	0.7360	0.9430	0.4307	0.5001	0.7764	0.5648	0.6619
0.5115	13550	2.9678	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5303	14050	3.2339	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5492	14550	2.7277	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5662	15000	2.7871	0.4401	0.6901	0.9219	0.5779	0.9194	0.7119	0.3740	0.7328	0.9353	0.4332	0.5234	0.7823	0.5642	0.6620
0.5681	15050	2.6724	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5870	15550	2.6279	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6058	16050	2.5004	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6228	16500	2.5688	0.4437	0.6924	0.9027	0.5736	0.9216	0.7172	0.3775	0.7315	0.9370	0.4319	0.5080	0.7848	0.5668	0.6607
0.6247	16550	2.7057	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6436	17050	2.4897	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6625	17550	2.8434	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6795	18000	2.7652	0.4569	0.7001	0.9046	0.5656	0.9260	0.7237	0.3724	0.7396	0.9416	0.4378	0.5192	0.7856	0.5731	0.6651
0.6813	18050	2.6545	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7002	18550	2.5231	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7191	19050	2.9636	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7361	19500	2.4418	0.4436	0.6939	0.9096	0.5608	0.9241	0.7248	0.3711	0.7414	0.9436	0.4408	0.5074	0.7771	0.5626	0.6616
0.7380	19550	2.8183	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7568	20050	2.7331	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7757	20550	2.7311	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7927	21000	2.5074	0.4596	0.6974	0.9254	0.5623	0.9271	0.7387	0.3742	0.7326	0.9443	0.4416	0.5160	0.7808	0.5712	0.6670
0.7946	21050	2.6516	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8135	21550	2.5811	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8323	22050	2.7768	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8493	22500	2.5435	0.4467	0.6969	0.9132	0.5656	0.9267	0.7322	0.3743	0.7298	0.9324	0.4412	0.5010	0.7791	0.5699	0.6622
0.8512	22550	2.5532	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8701	23050	2.4875	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8889	23550	2.6327	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9059	24000	3.1549	0.4503	0.6934	0.9130	0.5593	0.9270	0.7274	0.3757	0.7301	0.9316	0.4411	0.5135	0.7791	0.5659	0.6621
0.9078	24050	2.765	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9267	24550	3.0114	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9456	25050	2.8127	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9626	25500	2.4787	0.4515	0.6936	0.9221	0.5661	0.9269	0.7348	0.3706	0.7296	0.9333	0.4419	0.5100	0.7820	0.5667	0.6638
0.9644	25550	2.5551	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9833	26050	2.5297	-	-	-	-	-	-	-	-	-	-	-	-	-	-

Framework Versions

Python: 3.13.0
Sentence Transformers: 5.1.1
PyLate: 1.3.4
Transformers: 4.48.3
PyTorch: 2.6.0
Accelerate: 1.12.0
Datasets: 4.4.1
Tokenizers: 0.21.0

Citation

BibTeX

ColBERT-Zero

bibtex

@misc{chaffin2026colbertzeropretrainpretraincolbert,
  title         = {ColBERT-Zero: To Pre-train Or Not To Pre-train ColBERT models}, 
  author        = {Antoine Chaffin and Luca Arnaboldi and Amélie Chatelain and Florent Krzakala},
  year          = {2026},
  eprint        = {2602.16609},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url           = {https://arxiv.org/abs/2602.16609}, 
}

Sentence Transformers

bibtex

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084"
}

PyLate

bibtex

@inproceedings{DBLP:conf/cikm/ChaffinS25,
  author       = {Antoine Chaffin and
                  Rapha{"{e}}l Sourty},
  editor       = {Meeyoung Cha and
                  Chanyoung Park and
                  Noseong Park and
                  Carl Yang and
                  Senjuti Basu Roy and
                  Jessie Li and
                  Jaap Kamps and
                  Kijung Shin and
                  Bryan Hooi and
                  Lifang He},
  title        = {PyLate: Flexible Training and Retrieval for Late Interaction Models},
  booktitle    = {Proceedings of the 34th {ACM} International Conference on Information
                  and Knowledge Management, {CIKM} 2025, Seoul, Republic of Korea, November
                  10-14, 2025},
  pages        = {6334--6339},
  publisher    = {{ACM}},
  year         = {2025},
  url          = {https://github.com/lightonai/pylate},
  doi          = {10.1145/3746252.3761608},
}

Nomic Embed

bibtex

@article{DBLP:journals/tmlr/NussbaumMMD25,
  author       = {Zach Nussbaum and
                  John Xavier Morris and
                  Andriy Mulyar and
                  Brandon Duderstadt},
  title        = {Nomic Embed: Training a Reproducible Long Context Text Embedder},
  journal      = {Trans. Mach. Learn. Res.},
  volume       = {2025},
  year         = {2025},
  url          = {https://openreview.net/forum?id=IPmzyQSiQE},
  timestamp    = {Fri, 20 Jun 2025 14:19:48 +0200},
  biburl       = {https://dblp.org/rec/journals/tmlr/NussbaumMMD25.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

lightonai/ColBERT-Zero-supervised

ColBERT-Zero

Why ColBERT-Zero?

Controlled Comparison Design

The Three-Phase Training Pipeline

Phase 1 - Unsupervised Contrastive Pre-training

Phase 2 - Supervised Contrastive Fine-tuning

Phase 3 - Knowledge Distillation (KD)

Key Findings

1. The Standard Recipe Leaves Performance on the Table

2. Supervised + KD Is the Efficiency Sweet Spot

3. Prompt Alignment Is Non-Negotiable

Model Lineup

The Main Models (ColBERT-Zero)

The cheap-to-train ones (ModernColBERT-embed-base)

Intermediate Checkpoints

Full Performance on BEIR

Limitations & Discussion

Serving at Scale

Resources

Model Details

Model Description

Model Sources

Full Model Architecture

Usage

Retrieval

Indexing documents

Retrieving top-k documents for queries

Reranking

Evaluation

Metrics

Py Late Information Retrieval

Nano BEIR

Training Details

Training Datasets

medi_sts_wiki_rephrasal

fever_hn_mine

msmarco_distillation_simlm_rescored_reranked_min15

reddit_triples

medi_sts_flickr_sampled

nq_cocondensor_hn_mine_reranked_min15

nli_simcse_50negs_fixed

medi_supernli_sampled

hotpotqa_hn_mine_shuffled

medi_sts_stackexchange_dupe

Training Hyperparameters

Non-Default Hyperparameters

All Hyperparameters

Training Logs

Framework Versions

Citation

BibTeX

ColBERT-Zero

Sentence Transformers

PyLate

Nomic Embed

lightonai/
ColBERT-Zero-supervised