الباحثون

بيخون هو

المنشورات 2

نسخة أولية وصول مفتوح

Relevance Is Not Sufficient Evidence: Detecting Evidence Gaps Before Generation in RAG

Suting Chen, بيخون هو, يونمينغ شياو · 2026

Retrieval-augmented generation (RAG) grounds large language models in external sources, but retrieved passages often name the right entities without providing the facts needed to answer. Even when instructed to abstain, 12 generators answer 40.0-99.3% of insufficient-evidence questions. Training generators to abstain t …

نسخة أولية وصول مفتوح

Backdoor in the Loop: Compromising Agentic Search via Malicious Retrievers

Beining Xu, بيخون هو, يونمينغ شياو · 2026

Agentic retrieval-augmented generation (RAG) interleaves reasoning with repeated retrieval, giving the retriever influence over both the evidence an agent observes and its subsequent search decisions. We study retriever backdoors that exploit this feedback loop and repurpose weak backdoor purification to conceal their …

المؤلفون المشاركون