Metrics below are computed on small fixed subsets, not full benchmarks. Exact subset sizes: ambigqa=128, ifeval=64, mmlu=228. Reported values are from the final training step (step=1620, epoch=3).

Dataset

Hub dataset: sewon/ambig_qa (config_name=light)
Revision: 6e667596df70f17ba3c8e7be4b7361f6be8b60f8
Splits: train=train, validation=validation
Dev holdout: dev_fraction=0.1, split_seed=1729
Bootstrap resampling: False

System prompt

No system prompt (dataset.system_prompt: null).

Hyperparameters

Exact resolved settings used for this run (also attached as resolved_config.yaml):

yaml

model:
  name: Qwen/Qwen3.5-2B
  revision: main
  dtype: bfloat16
  enable_thinking: false
dataset:
  name: sewon/ambig_qa
  config_name: light
  revision: 6e667596df70f17ba3c8e7be4b7361f6be8b60f8
  train_split: train
  validation_split: validation
  dev_fraction: 0.1
  split_seed: 1729
  bootstrap: false
  max_train_examples: null
  max_dev_examples: null
  system_prompt_set: false
adaptation:
  method: full
training:
  num_train_epochs: 3.0
  max_steps: -1
  learning_rate: 2.0e-05
  weight_decay: 0.1
  warmup_ratio: 0.03
  per_device_train_batch_size: 4
  per_device_eval_batch_size: 16
  gradient_accumulation_steps: 8
  max_seq_length: 512
  logging_steps: 10
  save_every_steps: null
  save_total_limit: null
  gradient_checkpointing: true
  bf16: true
  tf32: true
  max_grad_norm: 1.0
  dataloader_num_workers: 2
  adam_beta1: 0.9
  adam_beta2: 0.999
  adam_epsilon: 1.0e-08
ensemble:
  size: 5
  base_seed: 42
  this_member_index: 0
  this_member_seed: 42
evaluation_subsets:
  ambigqa:
    subset_size: 128
    seed: 1001
  ifeval:
    subset_size: 64
    seed: 1002
  mmlu:
    subset_size: 228
    seed: 1004

Metrics breakdown

Final-step ensemble mean ± std (n=5)

AmbigQA accuracy: 0.0766 ± 0.0086
AmbigQA AlignScore: 0.1310 ± 0.0114
IFEval prompt_level_strict_accuracy: 0.4375 ± 0.0383
MMLU accuracy: 0.6061 ± 0.0114
train_loss: 1.1289 ± 0.0033
steps: 1620
epochs: 3

Per-member (final step)

member	seed	steps	epochs	train_loss	AmbigQA acc	AmbigQA AlignScore	IFEval strict	MMLU acc
0 (this repo)	42	1620	3	1.1297	0.0625	0.1165	0.4688	0.6184
1	1051	1620	3	1.1313	0.0781	0.1373	0.4688	0.5965
2	2060	1620	3	1.1237	0.0781	0.1422	0.4375	0.6140
3	3069	1620	3	1.1280	0.0859	0.1378	0.3750	0.5921
4	4078	1620	3	1.1320	0.0781	0.1210	0.4375	0.6096

Files

Model weights / adapter files from members/member_000/final/
resolved_config.yaml — full resolved training config
ensemble_metrics.png — train/eval curves for the whole ensemble