10 papers
Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning
Yanxi Chen, Weijie Shi, Yuexiang Xie +4
This work presents a general framework for training large language models (LLMs) to "Connect the Dots" (CoD), a meta-capability required by long-lifecycle agents: as an LLM-based A…
RL: Reflect-then-Retry Reinforcement Learning with Language-Guided Exploration, Pivotal Credit, and Positive Amplification
Weijie Shi, Yanxi Chen, Zexi Li +5
Reinforcement learning drives recent advances in LLM reasoning and agentic capabilities, yet current approaches struggle with both exploration and exploitation. Exploration suffers…
On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting
Wenhao Zhang, Yuexiang Xie, Yuchang Sun +5
Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are two prominent post-training paradigms for refining the capabilities and aligning the behavior of Large Language Mod…
Group-Relative REINFORCE Is Secretly an Off-Policy Algorithm: Demystifying Some Myths About GRPO and Its Friends
Chaorui Yao, Yanxi Chen, Yuchang Sun +5
Off-policy reinforcement learning (RL) for large language models (LLMs) is attracting growing interest, driven by practical constraints in real-world applications, the complexity o…
SeeUPO: Sequence-Level Agentic-RL with Convergence Guarantees
Tianyi Hu, Qingxu Fu, Yanxi Chen +2
Reinforcement learning (RL) has emerged as the predominant paradigm for training large language model (LLM)-based AI agents. However, existing backbone RL algorithms lack verified…
On the Entropy Dynamics in Reinforcement Fine-Tuning of Large Language Models
Shumin Wang, Yuexiang Xie, Wenhao Zhang +4
Entropy serves as a critical metric for measuring the diversity of outputs generated by large language models (LLMs), providing valuable insights into their exploration capabilitie…