4 citations · 4 across the 17 of their papers we have counts for
18 papers
World Model Science: Self-Organized Criticality, Weak Chaos, and Metastable Belief Dynamics in Long-Horizon LLM Agents
Xinyuan Song, Zekun Cai
Long-horizon LLM agents must maintain task state across extended sequences of observations, actions, tool calls, and intermediate beliefs. We study these trajectories through three…
MEMONDEMAND: A Memory Management System for Large-Scale Enterprise Data
Xinyuan Song, Bowen Zhu, Hasibul Haque +1
Enterprise repositories are large, heteroge- neous, and continuously updated, making re- trieval difficult when efficient access, source- faithful evidence, and cross-query adaptat…
MegaMem: A Retrieval Solution for Ultra-Large Context Windows
Xinyuan Song, Bowen Zhu, Hasibul Haque +1
Modern language models and agents increasingly require persistent memory for complete codebases, long interaction histories, and heterogeneous enterprise records. The key challenge…
HarnessMDP: Learning to Control LLM Agent Harnesses with Offline Reinforcement Learning
Haiwen Yi, Xinyuan Song
The execution harness is a critical component of LLM agents, shaping how the agent interacts with its environment and organizes multi-step execution. Prior work has shown that impr…
Measuring Harness-Induced Belief Divergence in Multi-Step LLM Agents
Haiwen Yi, Xinyuan Song
Software agents are becoming increasingly important in modern coding workflows, where they solve complex tasks through multi-step interaction with code repositories. A central comp…
Repair the Core, Not the Symptom: Planning-Graph Repair for Long-Horizon Agents
Xinyuan Song, Zekun Cai
Long-horizon language agents make interdependent decisions across multiple rounds, forming growing planning graphs in which later decisions depend on earlier states, observations,…
Ask the World Before Acting: Environment Probing for Calibrating Agent World Models
Xinyuan Song, Zekun Cai
Long-horizon language agents enable autonomous decision-making in complex environments by planning and executing sequences of interdependent actions. To act reliably, these agents…
Agentic vs. Parametric World Models: Hybrid Planning for Reliable Language Agents
Xinyuan Song, Zekun Cai
Language agents, such as ReAct, Reflexion, and RAP, implicitly predict how the world will change as they plan and act, using imagined state transitions to guide their decisions. Th…
Understanding Rollout Error in Graph World Models
Xinyuan Song, Zekun Cai
World models are increasingly used for planning by predicting future states under candidate actions. Many planning environments, however, are naturally graph-structured, with agent…
World-Model Collapse as a Phase Transition
Xinyuan Song, Zekun Cai
Water looks unchanged as it warms, then at a critical point it boils. We ask whether long-horizon language agents show an analogous transition in their implicit world models. In so…
AlgoBench: Benchmarking Algorithmic Adaptation in Code Generation
Xinyuan Song, Zekun Cai, Liang Zhao
High pass rates on established programming benchmarks such as HumanEval and LiveCodeBench do not always show whether a model can reason about algorithms. Many fixed benchmarks even…
AlgoSkill: Learning to Design Algorithms by Scheduling Human-Like Skills
Xinyuan Song, Zekun Cai, Liang Zhao
Designing an algorithm from a natural-language problem statement requires identifying the problem structure, reading constraints, choosing a suitable paradigm, checking correctness…
When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs
Xinyuan Song, Zekun Cai, Liang Zhao
Recursive self-training can degrade neural generative models when generated data is reused without fresh human data or external quality control. We study this risk in code LLMs, wh…
SkillOps: Managing LLM Agent Skill Libraries as Self-Maintaining Software Ecosystems
Hongji Pu, Xinyuan Song, Liang Zhao
Large language model agents increasingly rely on skill libraries for multi-step tasks, yet these libraries can accumulate persistent defects as skills are added, reused, patched, a…
Parameter-Efficient Temporal Domain Generalization
Yiheng Yao, Zekun Cai, Xinyuan Song +4
Temporal domain generalization (TDG) learns from a sequence of observed domains to predict models for unseen future domains. Existing parameter-dynamics methods become prohibitivel…
Learning to Recommend Multi-Agent Subgraphs from Calling Trees
Xinyuan Song, Liang Zhao
Multi-agent systems (MAS) increasingly solve complex tasks by orchestrating agents and tools selected from rapidly growing marketplaces. As these marketplaces expand, many candidat…
Learning-Based Automated Adversarial Red-Teaming for Robustness Evaluation of Large Language Models
Zhang Wei, Hanxuan Chen, Peilu Hu +19
Red-teaming is becoming a central part of large language model (LLM) safety evaluation, yet current practice still relies heavily on expert-written prompts or fixed benchmark suite…
StructPrune: Structured Global Pruning asymptotics with GPU Memory
Xinyuan Song, Guangji Bai, Liang Zhao
Pruning is critical for scaling large language models (LLMs). Global pruning achieves strong performance but requires memory, which is infeasible for billion-param…