arxivcs.CVcs.AI2026-07-01
LongVQUBench: Benchmarking Long-Term Video Quality Understanding of Vision-Language Models
Arpita Nema, Hanwei Zhu, Xi Zhang, Weisi Lin
The evaluation of long-term video quality understanding remains an open challenge for large vision-language models (LVLMs). Existing video quality benchmarks predominantly focus on short clips and isolated distortions, overlooking the temporal continuity, cumulative degradation,…