English
Related papers

Related papers: From Generic Correlation to Input-Specific Credit …

200 papers

On-policy reinforcement learning has become the dominant paradigm for reasoning alignment in large language models, yet its sparse, outcome-level rewards make token-level credit assignment notoriously difficult. On-Policy Distillation (OPD)…

Machine Learning · Computer Science 2026-04-14 Binbin Zheng , Xing Ma , Yiheng Liang , Jingqing Ruan , Xiaoliang Fu , Kepeng Lin , Benchang Zhu , Ke Zeng , Xunliang Cai

On-Policy Distillation (OPD) has gained wide attraction as an LLM post-training paradigm due to its effectiveness in improving capabilities without introducing model distribution drift, and consequently, regression in general tasks.…

Artificial Intelligence · Computer Science 2026-05-25 Aristotelis Lazaridis , Dylan Bates , Aman Sharma , Brian King , Vincent Lu , Jack FitzGerald

On-policy distillation (OPD) is an increasingly important paradigm for post-training language models. However, we identify a pervasive Scaling Law of Miscalibration: while OPD effectively improves task accuracy, it systematically traps…

Machine Learning · Computer Science 2026-04-21 Jiaxin Zhang , Xiangyu Peng , Qinglin Chen , Qinyuan Ye , Caiming Xiong , Chien-Sheng Wu

In settings where labeled verifiable training data is the binding constraint, each checked example should be allocated to the model and reward density where it is most informative. We identify a reward-density principle that governs this…

Machine Learning · Computer Science 2026-05-21 Yuanda Xu , Hejian Sang , Zhengze Zhou , Ran He , Zhipeng Wang , Alborz Geramifard

On-policy self-distillation (OPSD) is an emerging LLM post-training paradigm in which the model serves as its own teacher: conditioned on privileged information such as a reference trace or hint, the same policy provides dense token-level…

Machine Learning · Computer Science 2026-05-22 Hongbin Zhang , Chaozheng Wang , Kehai Chen , Youcheng Pan , Yang Xiang , Jinpeng Wang , Min Zhang

On-policy distillation (OPD) trains a student on its own trajectories with token-level teacher feedback and often outperforms off-policy distillation and standard reinforcement learning. However, we find that its standard advantage weighted…

Machine Learning · Computer Science 2026-05-14 Nan Jia , Haojin Yang , Xing Ma , Jiesong Lian , Shuailiang Zhang , Weipeng Zhang , Ke Zeng , Xunliang Cai , Zequn Sun

Knowledge distillation offers a promising path to transfer reasoning capabilities from large teacher models to efficient student models; however, existing token-level on-policy distillation methods require token-level alignment between the…

Computation and Language · Computer Science 2026-01-30 Jing Xiong , Hui Shen , Shansan Gong , Yuxin Cheng , Jianghan Shen , Chaofan Tao , Haochen Tan , Haoli Bai , Lifeng Shang , Ngai Wong

In this paper, we propose that small models may not need to absorb the cost of pre-training to reap its benefits. Instead, they can capitalize on the astonishing results achieved by modern, enormous models to a surprising degree. We observe…

Machine Learning · Computer Science 2024-05-06 Sean Farhat , Deming Chen

The success of RL for LLM post-training stems from an unreasonably uninformative source: a single bit of information per rollout as binary reward or preference label. At the other extreme, distillation offers dense supervision but requires…

Machine Learning · Computer Science 2026-02-12 Yuda Song , Lili Chen , Fahim Tajwar , Remi Munos , Deepak Pathak , J. Andrew Bagnell , Aarti Singh , Andrea Zanette

Knowledge distillation has attracted a great deal of interest recently to compress pre-trained language models. However, existing knowledge distillation methods suffer from two limitations. First, the student model simply imitates the…

Computation and Language · Computer Science 2023-05-18 Siyue Wu , Hongzhan Chen , Xiaojun Quan , Qifan Wang , Rui Wang

We consider the problem of imitation learning from a finite set of expert trajectories, without access to reinforcement signals. The classical approach of extracting the expert's reward function via inverse reinforcement learning, followed…

Machine Learning · Computer Science 2019-06-10 Ruohan Wang , Carlo Ciliberto , Pierluigi Amadori , Yiannis Demiris

Distilling reasoning traces from strong large language models into smaller ones is a promising route to improve intelligence in resource-constrained settings. Existing approaches face a fundamental trade-off: offline distillation from…

Computation and Language · Computer Science 2026-05-15 Yumeng Zhang , Zhengbang Yang , Yevin Nikhel Goonatilake , Zhuangdi Zhu

In this paper, we argue that mutual distillation between reinforcement learning policies serves as an implicit regularization, preventing them from overfitting to irrelevant features. We highlight two separate contributions: (i)…

Machine Learning · Computer Science 2025-09-25 Zhengpeng Xie , Jiahang Cao , Changwei Wang , Fan Yang , Marco Hutter , Qiang Zhang , Jianxiong Zhang , Renjing Xu

Learning from demonstrations in embodied control is often cast as behavioral cloning, and recent diffusion or flow-matching policies improve this paradigm by modeling multi-modal expert actions. Yet these methods remain offline supervised…

Machine Learning · Computer Science 2026-05-27 Zhenglin Wan , Jingxuan Wu , Xingrui Yu , Chubin Zhang , Mingcong Lei , Bo An , Ivor W. Tsang , Yang You

Self-distillation enables language models to learn on-policy from their own trajectories by using the same model as both student and teacher, with the teacher being conditioned on privileged information unavailable to the student. Such…

On-policy distillation (OPD) leverages dense teacher rewards to enhance reasoning models. However, scaling OPD to long-horizon tasks exposes a critical flaw: as the student's generated prefix inevitably diverges from the teacher's thought…

Machine Learning · Computer Science 2026-05-29 Zhicheng Yang , Zhijiang Guo , Yifan Song , Minrui Xu , Yongxin Wang , Yiwei Wang , Xiaodan Liang , Jing Tang

Reinforcement learning with verifiable rewards has become the standard recipe for improving LLM reasoning, but the dominant algorithm GRPO assigns a single trajectory-level advantage to every token, diluting the signal at pivotal reasoning…

Machine Learning · Computer Science 2026-05-25 Yu Li , Rui Miao , Tian Lan , Zhengling Qi

We introduce a novel approach to large language model (LLM) distillation by formulating it as a constrained reinforcement learning problem. While recent work has begun exploring the integration of task-specific rewards into distillation…

Machine Learning · Computer Science 2025-09-30 Matthieu Zimmer , Xiaotong Ji , Tu Nguyen , Haitham Bou Ammar

Self-Distillation Policy Optimization (SDPO) provides dense token-level credit assignment for reinforcement learning with large language models by leveraging the model's own feedback-conditioned predictions as a self-teacher. Unlike GRPO,…

Machine Learning · Computer Science 2026-05-28 Zehao Liu , Yuanpu Cao , Jinghui Chen , Vasant G. Honavar

Distillation efforts have led to language models that are more compact and efficient without serious drops in performance. The standard approach to distillation trains a student model against two objectives: a task-specific objective (e.g.,…

Computation and Language · Computer Science 2022-06-07 Zhengxuan Wu , Atticus Geiger , Josh Rozner , Elisa Kreiss , Hanson Lu , Thomas Icard , Christopher Potts , Noah D. Goodman