Researcher — paper.dinq.me

2026 顶会论文

作者位次 #4/7ICLR 2026Poster· Social Aspects / Fairness, Equity, Justice and Safety

Detecting Data Contamination from Reinforcement Learning Post-training for Large Language Models

Yongding Tao · Tian Wang · Yihong Dong · Huanyu Liu · Kechi Zhang · Hu XiaoLong · Ge Li

强化学习与后训练社会与人本安全OpenReview

作者位次 #4/14ACL 2026Poster· Natural Language Processing

RL - PLUS : Countering Capability Boundary Collapse of LLM s in Reinforcement Learning with Hybrid-policy Optimization

Yihong Dong · Xue Jiang · Yongding Tao · Huanyu Liu · Kechi Zhang · Lili Mou · Rongyu Cao · YINGWEI MA · Jue Chen · Binhua Li · Zhi Jin · Fei Huang · Yongbin Li · Ge Li

摘要

Reinforcement Learning with Verifiable Reward (RLVR) has significantly advanced the complex reasoning abilities of Large Language Models (LLMs). However, it struggles to break through the inherent capability boundaries of the base LLM, due to its essentially on-policy strategy coupled with LLM’s...

强化学习与后训练推理理论与学习动力学OpenReview 项目页

一作ACL 2026Findings· Findings of ACL

E vo C o T : Overcoming the Exploration Bottleneck in Reinforcement Learning for LLM s

Huanyu Liu · Jia Li · Yihong Dong · Chang Yu · Taozhi Chen · Lecheng Wang · Yongding Tao · Bin Gu · Ge Li

摘要

Reinforcement learning with verifiable reward (RLVR) has become a promising paradigm for post-training large language models (LLMs) to improve their reasoning capability. However, when the rollout accuracy is low on hard problems, the reward becomes sparse, limiting learning efficiency and causin...

强化学习与后训练推理多模态OpenReview 项目页

Huanyu Liu

获取完整研究档案

联系方式

2026 顶会论文

Detecting Data Contamination from Reinforcement Learning Post-training for Large Language Models

RL - PLUS : Countering Capability Boundary Collapse of LLM s in Reinforcement Learning with Hybrid-policy Optimization

E vo C o T : Overcoming the Exploration Bottleneck in Reinforcement Learning for LLM s

常见合作者