truncated-step-level-retrieval-reasoning

【效率与协作】Improve credit assignment in retrieval-augmented reasoning by truncating trajectories at single decision points. Generate k samples sharing a common prefix, differing only at the next step to isolate variation and reduce gradient variance by T-fold on T-step tasks.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准