Temporal Contrastive Learning for Video Temporal Reasoning in Large Vision-Language Models

TL;DR

This paper introduces TSADP, a novel framework that significantly improves temporal reasoning in large vision-language models by using dynamic prompts and contrastive learning, leading to better understanding of video sequences.

Contribution

The paper proposes TSADP, a new approach combining dynamic prompting and contrastive learning to enhance temporal reasoning in large vision-language models.

Findings

TSADP outperforms state-of-the-art models on VidSitu dataset

Improves tasks like entity association and temporal understanding

Human evaluations show better semantic coherence

Abstract

Temporal reasoning is a critical challenge in video-language understanding, as it requires models to align semantic concepts consistently across time. While existing large vision-language models (LVLMs) and large language models (LLMs) excel at static tasks, they struggle to capture dynamic interactions and temporal dependencies in video sequences. In this work, we propose Temporal Semantic Alignment via Dynamic Prompting (TSADP), a novel framework that enhances temporal reasoning capabilities through dynamic task-specific prompts and temporal contrastive learning. TSADP leverages a Dynamic Prompt Generator (DPG) to encode fine-grained temporal relationships and a Temporal Contrastive Loss (TCL) to align visual and textual embeddings across time. We evaluate our method on the VidSitu dataset, augmented with enriched temporal annotations, and demonstrate significant improvements over state-of-the-art models in tasks such as Intra-Video Entity Association, Temporal Relationship Understanding, and Chronology Prediction. Human evaluations further confirm TSADP's ability to generate coherent and semantically accurate descriptions. Our analysis highlights the robustness, efficiency, and practical utility of TSADP, making it a step forward in the field of video-language understanding.