arXiv:2608.22898v1 Announce Type: new
Abstract: Diffusion language models (DLMs) alleviate the inherent latency bottleneck of autoregressive (AR) large language models (LLMs), but their degraded gene...
By Hyeongsoo Lim, Jinyoung Kim, Eunseo Seo, Minho Jang, Jiwon Yoon
arXiv:2606. 00755v1 Announce Type: cross Abstract: Reinforcement learning from verifiable rewards improves the reasoning ability of large language models, but often suffers from entropy collapse, in which increasingly concentrated policies reduce rollout diversity and useful learning signals.
By Xuewei Yang, Jiachen Yu, Jie Wu, Shaoning Sun, Junjie Wang, Yujiu Yang
arXiv:2608. 14685v1 Announce Type: new Abstract: Knowledge distillation (KD) is widely used to transfer the capabilities of large language models (LLMs) to smaller students, but existing objectives often struggle to balance faithful imitation and robust generation.
By Shizhen Li, Zhiyu Shen, Yuyin Lu, Yunhe Pang, Jielin Song, Yanghui Rao, Fu Lee Wang
arXiv:2607. 08776v1 Announce Type: cross Abstract: Despite the success of knowledge distillation (KD) in Large Language Models (LLMs), the underlying mechanism behind its efficacy remains unclear.
By Qingzhuo Wang, Ruiyang Qin, Zhenxin Qin, Wen Shen, Zhihua Wei
The paper investigates how the choice of divergence in knowledge distillation affects the entropy of the student model. It shows that forward KL increases student entropy beyond the teacher’s, while reverse KL decreases it, and that interpolating between the two yields a smooth entropy change early in training but a sharp shift at convergence. The study also finds that on‑policy distillation’s lower entropy stems from token‑level reverse KL rather than sampling, positioning divergence as an implicit entropy regularizer, especially evident in self‑distillation scenarios.
By Nicolas Zucchet, Scott W. Linderman
arXiv:2606. 09456v1 Announce Type: new Abstract: On-Policy Distillation (OPD) has become a core technique in the post-training of Large Language Models (LLMs) for transferring knowledge from domain experts to student models.
By Yifan Niu, Han Xiao, Dongyi Liu, Zelong Wang, Dihong Gong, Yasheng Wang, Jia Li