puzzle-curriculum-grpo

【效率与协作】Post-train vision-language models using automatically-verifiable puzzle environments (Jigsaw, Rotation, PatchFit) with graded rewards. Implement exploration-aware curriculum combining difficulty weighting with solution-space diversity metrics. Track reasoning-answer consistency to prevent divergence during training.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准