28 papers
EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence
Qijin Zhang, Hanzhen Wang, Yuehan Zhao +4
Tool-using language model agents rely on environmental facts to decide whether an action is permitted. When a tool response omits a safety-relevant fact, an agent may carry out a p…
Bridging Exploration and Persona: Information-Directed Dueling for Test-Time Agent Personalization
Jie Mao, Pingchen Lu, Xiangyi Wang +4
Effective test-time personalization requires LLM agents to adapt to query-dependent user preferences from lightweight interaction. We study persona selection with binary comparison…
IF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent Systems
Jie Mao, Pingchen Lu, Zikun Qu +6
Improving every agent in an LLM-based multi-agent system can be prohibitively costly, making it important to identify which agent should be prioritized to achieve the greatest syst…
CB-Orchestrator: Adaptive Workflow Optimization for LLM Agents via Contextual Bandits
Jiahang Sun, Zhiwei Shang, Zhipiao Liu +4
Large Language Model (LLM) agents have demonstrated remarkable capabilities in tackling complex tasks through agentic workflows. However, manually designing these workflows is labo…
MESH-Harness: Self-Improving Agent Harnesses via Bandit-Guided Compositional Evolution
Zhiwei Shang, Yu Huo, Mingrong Gong +6
An agent harness is the code that organizes context, maintains state, and coordinates tool calls for a language model. We study how to improve the harness under a limited evaluatio…
Fusion is the New Mutation: Bandit-Guided Evolution on Workflow Graphs
Zhiwei Shang, Jiahang Sun, Mingrong Gong +9
Automated agentic workflow optimization relies on costly evaluations, making it essential to allocate a limited evaluation budget effectively. Multi-parent fusion can reuse designs…
GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution
Geyi Yang, Zikun Qu, Xiang Li +4
The executable harness surrounding a GUI model determines how observations are assembled, actions are executed, and verification, recovery, and termination are controlled. Compared…
SERA: Scale-Equalized Rollout Allocation for Maximum Likelihood Reinforcement Learning
Zihao Chen, Fanxiang Xiong, Hongran Ren +6
Maximum Likelihood Reinforcement Learning (MaxRL) targets prompt-wise log-success and has shown strong performance on reasoning tasks. Under finite rollout budgets, however, the es…
From Preference to Reciprocity: Decentralized Matching with Empirically Grounded LLM-agent Based Modeling
Wangxuan Fan, Xiaoyu Nie, Zhoutian Shi +5
Bipartite matching is a fundamental problem in game theory and market design. Classical approaches such as Gale--Shapley assume complete preferences and centralized computation, wh…
SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation
Zikun Qu, Min Zhang, Mingze Kong +5
On-policy distillation (OPD) provides dense teacher supervision on student-generated trajectories, but standard reverse-KL training can assign insufficient probability to other pla…
DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization
Jian Mu, Tianyi Lin, Chengwei Qin +2
Large language models are increasingly deployed in multi-turn interactive settings where users or environments can iteratively provide lightweight feedback. Unfortunately, optimizi…
UniScale: Adaptive Unified Inference Scaling via Online Joint Optimization of Model Routing and Test-Time Scaling
Kaiyu Huang, Xingyu Wang, Mingze Kong +6
In real-world deployments of large language models (LLMs), balancing inference quality and computational cost has become a central challenge. Existing approaches tackle this trade-…
Linear and Neural Dueling Bandits with Delayed Feedback
Xiangyi Wang, Pingchen Lu, Jie Mao +4
Contextual dueling bandits form a cornerstone of preference-based decision-making, with critical applications in recommender systems and large language model alignment. However, st…
ALSO: Adversarial Online Strategy Optimization for Social Agents
Xiang Li, Liping Yi, Mingze Kong +3
Social simulation provides a compelling testbed for studying social intelligence, where agents interact through multi-turn dialogues under evolving contexts and strategically adapt…
Why Zeroth-Order Adaptation May Forget Less: A Randomized Shaping Theory
Yao Shu, Jian Mu, Zhongxiang Dai
Continual learning requires new-task adaptation without damaging previously acquired capabilities. Recent forward-pass and zeroth-order (ZO) results show that low-query adaptation…
MASPOB: Bandit-Based Prompt Optimization for Multi-Agent Systems with Graph Neural Networks
Zhi Hong, Qian Zhang, Jiahang Sun +5
Large Language Models (LLMs) have achieved great success in many real-world applications, especially the one serving as the cognitive backbone of Multi-Agent Systems (MAS) to orche…
Workflow-R1: Group Sub-sequence Policy Optimization for Multi-turn Workflow Construction
Mingze Kong, Zikun Qu, Zhongquan Zhou +7
The rapid evolution of agentic workflows has demonstrated strong performance of LLM-based agents in addressing complex reasoning tasks. However, existing workflow optimization meth…
Real-Time Aligned Reward Model beyond Semantics
Zixuan Huang, Xin Xia, Yuxi Ren +10
Reinforcement Learning from Human Feedback (RLHF) is a pivotal technique for aligning large language models (LLMs) with human preferences, yet it is susceptible to reward overoptim…
Dual-Robust Cross-Domain Offline Reinforcement Learning Against Dynamics Shifts
Zhongjian Qiao, Rui Yang, Jiafei Lyu +5
Single-domain offline reinforcement learning (RL) often suffers from limited data coverage, while cross-domain offline RL handles this issue by leveraging additional data from othe…
Self-Reflective Generation at Test Time
Jian Mu, Qixin Zhang, Zhiyong Wang +5
Large language models (LLMs) increasingly solve complex reasoning tasks via long chain-of-thought, but their forward-only autoregressive generation process is fragile; early token…
FedPOB: Sample-Efficient Federated Prompt Optimization via Bandits
Pingchen Lu, Zhi Hong, Zhiwei Shang +6
The performance of large language models (LLMs) is highly sensitive to the input prompt, making prompt optimization a critical task. However, real-world application is hindered by…
T-POP: Test-Time Personalization with Online Preference Feedback
Zikun Qu, Min Zhang, Mingze Kong +7
Personalizing large language models (LLMs) to individual user preferences is a critical step beyond generating generically helpful responses. However, current personalization metho…
Adaptive Batch-Wise Sample Scheduling for Direct Preference Optimization
Zixuan Huang, Yikun Ban, Lean Fu +4
Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences. However, its performance is highly depen…
Convergence Rates of Constrained Expected Improvement
Haowei Wang, Jingyi Wang, Zhongxiang Dai +3
Constrained Bayesian optimization (CBO) methods have seen significant success in black-box optimization with constraints. One of the most commonly used CBO methods is the constrain…