arXiv:2608. 09447v1 Announce Type: cross Abstract: On-policy distillation (OPD) aligns a student with a teacher on trajectories sampled from the student itself, reducing the train-test state mismatch of offline distillation.
By Zehao Chen, Gongxun Li, Tianxiang Ai, Yifei Li, Zixuan Huang, Wang Zhou, Tao Huang, Fuzhen Zhuang, Xianglong Liu, Jianxin Li, Deqing Wang, Yikun Ban
The paper investigates why reinforcement learning with verifiable rewards (RLVR) reduces the diversity of solutions in reasoning tasks. By analyzing the Countdown task, the authors show that RLVR contracts the solution space mainly at the entrance—before the first arithmetic operation—causing a 67% drop in solution coverage. They demonstrate that providing an unselected entrance prefix or applying entrance‑targeted interventions can restore or even improve coverage without harming accuracy.
By Qiancheng Zhou, Ruizhe Li
The study measured the impact of a single training example on a GPT‑2 model by running 24 counterfactual experiments. 32 models were trained from scratch on OpenWebText, and at a specific training step a single batch row was replaced with a 194‑token passage under three conditions (fluent prose, fabricated subject, random characters) or left unchanged. Results showed that the passage was learned from one exposure and decayed, with measurable differences in cross‑entropy up to 50 steps after injection but no lasting effect at the final step.
By Zachary Speck, Asa Shepard
The paper introduces Selective Supervision for Direct-OPD (S$^2$D-OPD), a refinement of Direct On-Policy Distillation that filters out states where the teacher’s policy change is minimal, as measured by the teacher‑reference Jensen‑Shannon divergence. By masking low‑divergence states and keeping only the top 10% of states per response, S$^2$D-OPD improves held‑out accuracy on AIME and HMMT benchmarks across multiple teacher‑student pairs without additional forward passes.
By Yibo Zhao, Zixuan Yang, Yunshi Lan, Xiang Li
The paper introduces Circuit Reasoning Score (CRS), a data‑selection signal for reinforcement learning with verifiable rewards that uses attention‑head activity from a frozen base model to gauge reasoning engagement. CRS is computed in a single forward pass without reward labels or rollouts, and it shows that selecting problems with the lowest reasoning‑circuit engagement can outperform random selection on several medium‑difficulty benchmarks. However, the benefit depends on domain, model scale, and reward conditions, indicating that data selection in this setting is regime‑dependent rather than a fixed ranking of problem quality.
By Zhuofan Chen, Ziqian Jiao, Yikai Cui, Zhixin Cai, Jun Bai, Wenge Rong
arXiv:2610.02391v1 Announce Type: new
Abstract: When a large language model solves a mathematical problem, its reasoning is largely hierarchical, and the solution often branches at a few tokens where...
By Zeyong Zhang, Tung Sum Thomas Kwok, Tengfei Ma, Mengjia Xu