Causality Matters: How Temporal Information Emerges in Video Language Models
By: Yumeng Shi , Quanyu Long , Yin Wu and more
Potential Business Impact:
Lets computers understand video time without special time codes.
Video language models (VideoLMs) have made significant progress in multimodal understanding. However, temporal understanding, which involves identifying event order, duration, and relationships across time, still remains a core challenge. Prior works emphasize positional encodings (PEs) as a key mechanism for encoding temporal structure. Surprisingly, we find that removing or modifying PEs in video inputs yields minimal degradation in the performance of temporal understanding. In contrast, reversing the frame sequence while preserving the original PEs causes a substantial drop. To explain this behavior, we conduct substantial analysis experiments to trace how temporal information is integrated within the model. We uncover a causal information pathway: temporal cues are progressively synthesized through inter-frame attention, aggregated in the final frame, and subsequently integrated into the query tokens. This emergent mechanism shows that temporal reasoning emerges from inter-visual token interactions under the constraints of causal attention, which implicitly encodes temporal structure. Based on these insights, we propose two efficiency-oriented strategies: staged cross-modal attention and a temporal exit mechanism for early token truncation. Experiments on two benchmarks validate the effectiveness of both approaches. To the best of our knowledge, this is the first work to systematically investigate video temporal understanding in VideoLMs, offering insights for future model improvement.
Similar Papers
Improving Temporal Understanding Logic Consistency in Video-Language Models via Attention Enhancement
CV and Pattern Recognition
Fixes videos so computers understand them better.
An Empirical Study on How Video-LLMs Answer Video Questions
CV and Pattern Recognition
Explains how AI understands videos to make them faster.
A Matter of Time: Revealing the Structure of Time in Vision-Language Models
CV and Pattern Recognition
Lets computers understand when pictures were taken.