9 papers
When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs
Jiayi Li, Ruizhe Li
Before invoking external tools, an agentic LLM must select among a K-way action space: executing a call, seeking clarification, answering directly, or declining. While internal act…
When Uncertainty Isn't Enough: An Empirical Study of Self-Correction in Code Generation
Pranav Rakasi, Maanas Lalwani, Arnav Srivastava +4
Large language models for code generation often produce incorrect solutions without reliable indicators of failure. We study whether uncertainty estimation methods developed for na…
Crayotter: Traceable Multi-Agent Workflows for Long-Form Video Editing
Lecheng Yan, Yichong Zhang, Xiantao Xu +12
Long-form video editing over heterogeneous footage requires agents to coordinate source selection, multimodal analysis, timeline construction, narration and subtitle alignment, ren…
Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation
Uwe König, Hamza Kazmi, Ruizhe Li +1
Distillation of a language model intended to transfer benign behavior to a student model may also transfer undesirable characteristics, if they are present in the teacher model, a…
Preference Optimization Drives Monoculture in LLM Prediction Markets
James Begin, Brendan Gho, Suman Muppavarapu +6
Prediction markets rest on the independence of participant errors. As LLM agents become active traders on platforms like Kalshi and Polymarket, we ask: does this independence hold…
Visuals Lie, Consistency Speaks: Disentangling Spatial Attention from Reliability in Vision-Language Models
Logan Mann, Yi Xia, Ajit Saravanan +6
Multimodal Foundation Models are increasingly used as reasoning agents, making reliability, knowing when a model may hallucinate, critical. A common intuition, which we call the At…
Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback
Lecheng Yan, Ruizhe Li, Xicheng Han +5
Tool-using LLM agents increasingly rely on external tools to make consequential decisions, yet most existing agent-security benchmarks and defenses implicitly assume that tool feed…
Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs
Lecheng Yan, Ruizhe Li, Guanhua Chen +5
Reinforcement Learning with Verifiable Rewards (RLVR) is highly effective for enhancing LLM reasoning, yet recent evidence shows models like Qwen 2.5 achieve significant gains even…
Marco-Bench-MIF: On Multilingual Instruction-Following Capability of Large Language Models
Bo Zeng, Chenyang Lyu, Sinuo Liu +14
Instruction-following capability has become a major ability to be evaluated for Large Language Models (LLMs). However, existing datasets, such as IFEval, are either predominantly m…