Ming Fan

The paper introduces EXHIB, a comprehensive benchmark of five real-world datasets, to evaluate Function Similarity Detection, demonstrating that current models fail to generalize across diverse low- and high-level binary variations.

The paper introduces BadStyle, a novel backdoor attack framework that generates natural, stealthy poisoned samples using LLMs to compromise various LLMs with high success rates and robust activation.

The paper introduces LITMUS, a novel benchmark that rigorously tests LLM agents for dangerous, physical-layer behavioral jailbreaks in real OS environments, revealing that current agents frequently execute high-risk operations despite safety guardrails.

The paper argues that current search agents often verify existing knowledge rather than genuinely searching, and introduces LiveBrowseComp, a new benchmark to measure true evidence-driven discovery.

This paper proposes Tail-Aware Adaptive-k (TAA-k), a training-free framework for adaptive context selection in retrieval-augmented generation systems using Extreme Value Theory.