12 papers
Beyond Retrieval Failure: Understanding Why LLMs Miss Evidence in Multi-Hop RAG
Jiaqi Wei, Wenxuan Huang, Chengdong Xu +2
Multi-hop retrieval-augmented generation (RAG) requires an agent to retrieve evidence for a sequence of dependent sub-questions. When a wrong answer lacks a required passage, the p…
AgentPatch: Coarse-to-Fine Weak-Task Repair for Merging Agentic Multimodal Large Language Models
Zibo Shao, Baochen Xiong, Chengdong Xu +6
Agentic multimodal large language models (MLLMs) extend multimodal perception and reasoning with planning, tool use, and interaction in dynamic environments. Yet current models are…
Efficient Online LLM Watermark Detection via Rao-Blackwellized E-Processes
Lu Luo, Dandan Mo, Chengdong Xu +4
As large language models (LLMs) are increasingly deployed, reliable and efficient mechanisms for distinguishing AI-generated text from human-written content have become essential.…
UCOB: Learning to Utilize and Evolve Agentic Skills via Credit-Aware On-Policy Bidirectional Self-Distillation
Songjun Tu, Chengdong Xu, Qichao Zhang +6
Skill memories can improve agentic reinforcement learning by reusing past experience as textual guidance, but retrieved skills are not oracular: they may help in one state while mi…
MotionHalluc: Diagnosing Kinematic Hallucinations in Fine-Grained Motion Reasoning
Weile Guo, Shenghong He, Danying Mo +3
Motion instruction generation in cross-video comparison aims to produce corrective feedback that describes the differences between a query and a reference motion. However, existing…
Adaptive Coarse-to-Fine Subgoal Refinement for Long-Horizon Offline Goal-Conditioned Reinforcement Learning
Kaiqiang Ke, Shenghong He, Chengdong Xu +3
Offline goal-conditioned reinforcement learning (GCRL) is challenging in long-horizon tasks, where distant state--goal pairs provide weak supervision and value estimates become vul…
TacoMAS: Test-Time Co-Evolution of Topology and Capability in LLM-based Multi-Agent Systems
Chen Xu, Yicheng Hu, Ruizi Wang +4
Multi-agent systems (MAS) have emerged as a promising paradigm for solving complex tasks. Recent work has explored self-evolving MAS that automatically optimize agent capabilities…
EvoMAS: Learning Execution-Time Workflows for Multi-Agent Systems
Chengdong Xu, Kaiqiang Ke, Ziheng Liu +4
Large language model (LLM)-based multi-agent systems have shown strong potential on complex tasks through agent specialization, tool use, and collaborative reasoning. However, most…
AndroTMem: From Interaction Trajectories to Anchored Memory in Long-Horizon GUI Agents
Yibo Shi, Jungang Li, Linghao Zhang +25
Long-horizon GUI agents are a key step toward real-world deployment, yet effective interaction memory under prevailing paradigms remains under-explored. Replaying full interaction…
Dynamic Dual-Granularity Skill Bank for Agentic RL
Songjun Tu, Chengdong Xu, Qichao Zhang +5
Agentic RL can benefit substantially from reusable experience, yet existing skill-based methods mainly extract trajectory-level guidance and often lack principled mechanisms for ma…
Context-Picker: Dynamic context selection using multi-stage reinforcement learning
Siyuan Zhu, Chengdong Xu, Kaiqiang Ke +1
In long-context question answering, selecting the appropriate scope of context for a query remains a key and unresolved challenge. Insufficient context can lead to missing essentia…
HR: Hierarchical Hindsight Reflection for Multi-Task LLM Agents
Shicheng Ye, Chao Yu, Kaiqiang Ke +2
Large language model (LLM)-based agents have shown strong potential in multi-task scenarios, owing to their ability to transfer knowledge across diverse tasks. However, existing ap…