15 papers
QREF: Improving DeepSearch Agents by Aligning Search with Reasoning via Query Refinement and Evidence Filtering
Yujia Chen, Yuhui Wang, Junzhe Wang +15
DeepSearch agents achieve strong performance on complex information-seeking tasks through iterative search and reasoning. However, they often issue intent-misaligned, under-constra…
ReDraft, Don't Just Distill: Reference-Driven Revision for Continual VLLM Post-Training
Zhihao Zhang, Mingqi Wu, Qiaole Dong +15
Continual post-training of large multimodal models should add new capabilities while preserving those from pre-training, and the two goals pull in opposite directions. SFT gives ex…
Sci-MMR: Benchmarking Multi-Step Evidence-Grounded Scientific Reasoning in Multimodal Agents
Jiaqiang Li, Yajie Yang, Zhiheng Xi +15
Autonomous research agents are increasingly expected to search the literature, analyze experimental evidence, and generate scientific hypotheses. These capabilities require multi-s…
SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information
Junjie Ye, Zhuohui Sheng, Shaofan Liu +12
Large language models (LLMs) are increasingly deployed as mobile assistants, where a key challenge is leveraging personal information scattered across multiple applications (apps)…
IACM-RL: Intent-Aware Context Management and Reinforcement Learning for Complex Tool Invocation under Dynamic Intent Fluctuations
Dingwei Zhu, Jiahan Li, Chengjun Pan +22
Executing long-horizon tool invocations in real-world environments is severely challenged by dynamic user intent noise. Existing methods attempt robustness via implicit history sca…
AgentGym2: Benchmarking Large Language Model Agents in De-Idealized Real-World Environments
Zhiheng Xi, Dingwen Yang, Jiaqi Liu +21
Language agents, i.e., LLM agents, progress rapidly and are increasingly deployed in production environments. This trend underscores the urgent need for rigorous and realistic eval…
Entropy Polarity in Reinforcement Fine-Tuning: Direction, Asymmetry, and Control
Jiazheng Zhang, Ziche Fu, Junrui Shen +17
Policy entropy has emerged as a fundamental measure for understanding and controlling exploration in reinforcement learning with verifiable rewards (RLVR) for LLMs. However, existi…
Prefix-Adaptive Block Diffusion for Efficient Document Recognition
Mingxu Chai, Ziyu Shen, Chenyu Liu +3
Block Diffusion Models (BDMs) support parallel generation, flexible-length output, and KV caching, making them promising for efficient document parsing. However, existing BDMs bind…
EVPO: Explained Variance Policy Optimization for Adaptive Critic Utilization in LLM Post-Training
Chengjun Pan, Shichun Liu, Jiahang Lin +10
Reinforcement learning (RL) for LLM post-training faces a fundamental design choice: whether to use a learned critic as a baseline for policy optimization. Classical theory favors…
AgentV-RL: Scaling Reward Modeling with Agentic Verifier
Jiazheng Zhang, Ziche Fu, Zhiheng Xi +13
Verifiers have been demonstrated to enhance LLM reasoning via test-time scaling (TTS). Yet, they face significant challenges in complex domains. Error propagation from incorrect in…
JFTA-Bench: Evaluate LLM's Ability of Tracking and Analyzing Malfunctions Using Fault Trees
Yuhui Wang, Zhixiong Yang, Ming Zhang +10
In the maintenance of complex systems, fault trees are used to locate problems and provide targeted solutions. To enable fault trees stored as images to be directly processed by la…
MagicAgent: Towards Generalized Agent Planning
Xuhui Ren, Shaokang Dong, Chen Yang +21
The evolution of Large Language Models (LLMs) from passive text processors to autonomous agents has established planning as a core component of modern intelligence. However, achiev…
DFPO: Scaling Value Modeling via Distributional Flow towards Robust and Generalizable LLM Post-Training
Dingwei Zhu, Zhiheng Xi, Shihan Dou +17
Training reinforcement learning (RL) systems in real-world environments remains challenging due to noisy supervision and poor out-of-domain (OOD) generalization, especially in LLM…
DVPO: Distributional Value Modeling-based Policy Optimization for LLM Post-Training
Dingwei Zhu, Zhiheng Xi, Shihan Dou +15
Reinforcement learning (RL) has shown strong performance in LLM post-training, but real-world deployment often involves noisy or incomplete supervision. In such settings, complex a…
VRPO: Rethinking Value Modeling for Robust RL under Noisy Supervision in LLM Post-Training
Dingwei Zhu, Shihan Dou, Zhiheng Xi +16
Reinforcement Learning (RL) in real-world environments often suffers from ambiguous or incomplete reward supervision, which undermines policy stability and generalization. Such noi…