15 papers
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…
Learning to Control LLM Agent Harnesses with Offline Reinforcement Learning
Haiwen Yi, Xinyuan Song
Large language model (LLM) agents are usually improved by changing prompts, models, or hand-written workflows, while the execution harness around the model is treated as fixed infr…
Measuring Harness-Induced Belief Divergence in Multi-Step LLM Agents
Haiwen Yi, Xinyuan Song
Software-agent benchmarks usually report whether an agent solves a task, but the agent reaches that outcome through a harness that controls what it sees, which actions it can take,…
Repair the Amplifier, Not the Symptom: Stable World-Model Correction for Agent Rollouts
Xinyuan Song, Zekun Cai
Long-horizon language agents increasingly maintain executable world models in the form of planning graphs, where tool calls, validators, memory updates, recovery branches, and fina…
Ask the World Before Acting: Environment Probing for Calibrated Agent World Models
Xinyuan Song, Zekun Cai
Language agents acting over long horizons must maintain beliefs about tool states, object locations, graph edges, and subgoal dependencies. When these beliefs drift, failures can b…
Agent vs. Parametric World Models: Hybrid Planning for Reliable Language Agents
Xinyuan Song, Zekun Cai
Language agents plan by generating not only actions but also implicit predictions of how the world will change. These imagined state updates make agents flexible, but they also cre…