The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

olmo32b — MATH-500 eval rollouts

MATH-500 evaluation rollouts for the Olmo-3-32B GRPO run (base allenai/Olmo-3-1125-32B, RL-Zero prompt, LoRA r=64, 300 steps). Adapters: ReasoningRegisters/olmo32b.

Layout (same as qwen14b-math500-eval-rollouts):

  • vanilla/step{0,50,100,150,200,250,300}/ — plain prompt, 500 problems × 8 rollouts, T=0.6, top-p 0.95, 8,192-token cap. step0 = base model; step50+ scored on merged weights.
  • vanilla/step300_roll32/ — final row: 500 × 32 rollouts, 16,384-token cap.
  • cue/step0/ — base model with the forced opener \n\nOkay, in the prompt (the condition the cue-forced run trains in). Later cue steps land as that run's checkpoints are evaluated.

Each directory holds 8 shard_XX.json (per-problem counts: n, c_correct, c_commit, c_capped) and 8 text_shard_XX.jsonl (one line per rollout: problem_key, rollout, gt, correct, capped, n_tokens, full text).

Headline numbers: vanilla pass@1 0.6747 (step0) → 0.9167 (step300); final row 0.9268; cue step0 0.8518.

Downloads last month
103