Co-trained with: Qwen2.5-3B x Qwen2.5-7B (weak-to-strong size pair) (this repo is group A)
Training data: q1716523669/MATH-Level345 (8,740 problems, MATH levels 3-5)
Eval: MATH-500 (500 prompts, pass@1, temperature 0.6)
This checkpoint: step 110 of 136 (2 epochs) — the best one
Its score: 0.6820

best is the step with the highest eval; end is step 136. Both are published because they answer different questions and the gap between them is a result in itself. The sibling repo is cogrpo-w2s-qwen25-3b-x-qwen25-7b-math345-groupA-qwen25-3b-end.

Hyperparameters

Effective batch 128 prompts per optimizer step, lr 3e-6, 2 epochs, 12 generations per prompt, 3072-token completions, temperature 1.0 for rollouts and 0.6 for eval, beta 0, bnpo loss, scale_rewards group, seed 42.

Eval curve

step	MATH-500 pass@1
0	0.5360
10	0.5020
20	0.6180
30	0.6300
40	0.6360
50	0.6480
60	0.6340
70	0.6500
80	0.6360
90	0.6540
100	0.6280
110	0.6820
120	0.6520
130	0.6420

Files for audit

train.log — the complete training log this checkpoint came from
eval_curve.csv — the table above, machine-readable
run_config.json — resolved config as the trainer saw it

Code: williamium3000/trl-projects, branch n3-interaction-modes, projects/co-grpo-dp/.

q1716523669/cogrpo-w2s-qwen25-3b-x-qwen25-7b-math345-groupA-qwen25-3b-best

cogrpo-w2s-qwen25-3b-x-qwen25-7b-math345-groupA-qwen25-3b · best checkpoint (step 110)

Hyperparameters

Eval curve

Files for audit

q1716523669/
cogrpo-w2s-qwen25-3b-x-qwen25-7b-math345-groupA-qwen25-3b-best