14 papers
Environment Evolution for Terminal Agents
Zhiyuan Fan, Tinghao Yu, Yuanjun Cai +9
Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less…
Distill Where You Fail: Recovering Learning Signals of Negative RL-Groups from Adaptive Teacher Guidance
Zhuowen Han, Jinwei Xiao, Zhengxi Lu +9
Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for post-training large language models (LLMs). While Group Relative Policy Optimization (GRPO)…
SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution
Zhiyuan Yao, Yuxin Chen, Zhengxi Lu +13
Large language model agents often encounter related yet distinct tasks that share reusable solution patterns. Yet standard agentic reinforcement learning treats tasks as independen…
Finding the Evidence: Discovering Decision-Supporting Tokens for On-Policy Reasoning Distillation
Jinwei Xiao, Zhuowen Han, Yueqing Sun +6
On-policy distillation transfers reasoning ability through dense token-level supervision, yet the nature of the transferable signal remains unclear. We discover that reasoning chai…
Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments
Yuxin Chen, Xiaodong Cai, Junfeng Fang +9
Recent advances in large language models (LLMs) have facilitated the widespread deployment of LLMs as interactive agents capable of reasoning, planning, and tool use. Despite stron…
Emotional intelligence in large language models is fragmented across perception, cognition, and interaction
Minghao Lv, Lu Chen, Enchang Zhang +6
As large language models (LLMs) are increasingly integrated into emotionally sensitive domains, the structural integrity of their emotional intelligence (EI) becomes a critical fro…
Self-Distilled Agentic Reinforcement Learning
Zhengxi Lu, Zhiyuan Yao, Zhuowen Han +8
Reinforcement learning (RL) has emerged as a central paradigm for post-training LLM agents, yet its trajectory-level reward signal provides only coarse supervision for long-horizon…
MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning
Yuxin Liu, Ziang Ye, Yueqing Sun +6
Current interactive LLM agents rely on goal-conditioned stepwise planning, where environmental understanding is acquired reactively during execution rather than established beforeh…
Toward Scalable Terminal Task Synthesis via Skill Graphs
Zhiyuan Fan, Tinghao Yu, Yuanjun Cai +8
Terminal agents have demonstrated strong potential for autonomous command-line execution, yet their training remains constrained by the scarcity of high-quality and diverse executi…
Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Models
Dan Shi, Zhuowen Han, Simon Ostermann +3
Reinforcement learning (RL)-based post-training often improves the reasoning performance of large language models (LLMs) beyond the training domain, while supervised fine-tuning (S…
DEP: A Decentralized Large Language Model Evaluation Protocol
Jianxiang Peng, Junhao Li, Hongxiang Wang +15
With the rapid development of Large Language Models (LLMs), a large number of benchmarks have been proposed. However, most benchmarks lack unified evaluation standard and require t…
DR-LoRA: Dynamic Rank LoRA for Fine-Tuning Mixture-of-Experts Models
Guanzhi Deng, Bo Li, Ronghao Chen +5
Mixture-of-Experts (MoE) has become a prominent paradigm for scaling Large Language Models (LLMs). Parameter-efficient fine-tuning methods, such as LoRA, are widely adopted to adap…
Revisiting Entropy in Reinforcement Learning for Large Reasoning Models
Renren Jin, Pengzhi Gao, Yuqi Ren +6
Reinforcement learning with verifiable rewards (RLVR) has emerged as a prominent paradigm for enhancing the reasoning capabilities of large language models (LLMs). However, the ent…
From Curated Data to Scalable Models: Continual Pre-training of Dense and MoE Large Language Models for Tibetan
Lei Yang, Leiyu Pan, Bojian Xiong +14
Large language models (LLMs) have achieved remarkable success across a wide range of natural language processing tasks, yet their performance remains heavily biased toward high-res…