Breaking Down Video LLM Benchmarks: Knowledge, Spatial Perception, or True Temporal Understanding?
This work addresses the need for more accurate evaluation of video large language models (LLMs) by identifying and mitigating limitations in current benchmarks, which is incremental but important for researchers and developers in video AI.
The authors tackled the problem that existing video understanding benchmarks conflate knowledge-based and image-based questions, obscuring true temporal reasoning ability, by proposing VBenchComp, an automated pipeline that categorizes questions into domains like LLM-Answerable, Semantic, and Temporal, revealing nuanced model weaknesses hidden by traditional scores.
Existing video understanding benchmarks often conflate knowledge-based and purely image-based questions, rather than clearly isolating a model's temporal reasoning ability, which is the key aspect that distinguishes video understanding from other modalities. We identify two major limitations that obscure whether higher scores truly indicate stronger understanding of the dynamic content in videos: (1) strong language priors, where models can answer questions without watching the video; and (2) shuffling invariance, where models maintain similar performance on certain questions even when video frames are temporally shuffled. To alleviate these issues, we propose VBenchComp, an automated pipeline that categorizes questions into different domains: LLM-Answerable, Semantic, and Temporal. Specifically, LLM-Answerable questions can be answered without viewing the video; Semantic questions remain answerable even when the video frames are shuffled; and Temporal questions require understanding the correct temporal order of frames. The rest of the questions are labeled as Others. This can enable fine-grained evaluation of different capabilities of a video LLM. Our analysis reveals nuanced model weaknesses that are hidden by traditional overall scores, and we offer insights and recommendations for designing future benchmarks that more accurately assess video LLMs.