AI RESEARCH
INFACT: A Diagnostic Benchmark for Induced Faithfulness and Factuality Hallucinations in Video-LLMs
arXiv CS.AI
•
ArXi:2603.11481v1 Announce Type: cross Despite rapid progress, Video Large Language Models (Video-LLMs) remain unreliable due to hallucinations, which are outputs that contradict either video evidence (faithfulness) or verifiable world knowledge (factuality). Existing benchmarks provide limited coverage of factuality hallucinations and predominantly evaluate models only in clean settings. We