Run ID: kanha.ai-0.6b-grounded-qlora-3ep-v1
Base model: Qwen/Qwen3-0.6B
Base model revision: c1899de289a04d12100db370d81485cdf75e47ca
Tokenizer revision: c1899de289a04d12100db370d81485cdf75e47ca
Training method: qlora
Final merged dtype: bfloat16
Source site: https://kanha.ai
Dataset hash: cf1aa836fc54b175ded08dae5101a741f751aee763b60f92fd49b0c038bee88a
Train split: 147 records (021e5b27f59e23d59634d0dbef221b61e43c288fa6a6e03b210d775b5efd8a1f)
Validation split: 21 records (4a82a1d0a25216d926ff54cd10246b80f147aaa021c4cd45417da02b4a76f01d)
Holdout split: 0 records (e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855)

Hyperparameters

Maximum sequence length: 4096
Seed: 42
Epochs: 3.0
Learning rate: 0.0001
Per-device batch size: 8
Gradient accumulation steps: 2
Warmup ratio: 0.05
Assistant-only loss: true
LoRA rank: 16
LoRA alpha: 16
LoRA dropout: 0.05
LoRA targets: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Evaluation

Model quality was evaluated with oracle source context. Retrieval quality is a separate gate recorded in research/retrieval-metrics.json.

dates_recall: 1.0
deterministic_pass_rate: 0.38461538461538464
list_recall: 0.6083333333333333
numbers_recall: 0.9641025641025641
refusal_rate: 0.038461538461538464
requires_review_rate: 0.0
total: 26
unsupported_value_rate: 0.0
urls_recall: 1.0

Deterministic scoring and the server-side Transformers benchmark are not browser qualification. Validate the exact converted model in the target browser and device environments.

MLC availability

MLC artifacts using q4f16_1 quantization are available under mlc/.

Grounded inference contract

Inference requires retrieved source context. A bare question without retrieved source context is outside the trained and evaluated contract.

Prompt contract identity: d0dcb1de0b9d601267a3cbf39420b1f0a2c5749d80b96669a2e9fcf19a280bfb
Chat formatting: The model's native chat template is used with thinking disabled (enable_thinking=False).

System prompt (exact):

Answer only from the supplied context. Be concise. If the answer is absent from the context, respond exactly: I can't answer that from the provided context.

User template (exact):

text

Context:
<context>

Question:
<question>

Refusal string (exact):

I can't answer that from the provided context.

Intended use

This checkpoint is intended for research comparing training methods on the same Kanha website-derived dataset and for controlled evaluation of website question answering.

Limitations

The checkpoint can produce incorrect, incomplete, or stale answers. It may memorize training content. Review outputs, test representative failure cases, and qualify the exact runtime before any user-facing use.

Provenance artifacts

research/run-manifest.json
research/training-config.yaml
research/publication-inventory.json
research/conversion-manifest.json
research/evaluation/metrics.json
research/evaluation/evaluation-manifest.json

Warlord-K/kanha-kanha.ai-0.6b-grounded-qlora-3ep-v1

Kanha Qwen3 experiment

Run identity