作者位次 #2/6SIGIR 2026Oral· Information Retrieval (Full)
Mitigating Evidence Suppression: Bi-level Active Evidence Injection for Educational Video Understanding
Cheng Liu · Yiping Wang · Quanlong Guan · Chaobo He · Xingyu Zhu · Liangda Fang
摘要
Large Vision--Language Models (LVLMs) frequently fail on knowledge-intensive educational video QA despite the presence of requisite visual evidence. Through region-level analysis, we identify a systematic evidence-suppression pattern: task-critical tokens (e.g., diagrams) exhibit lower representa...