Efficient Temporal Extrapolation of Multimodal Large Language Models with Temporal Grounding Bridge

TL;DR

This paper introduces the Temporal Grounding Bridge (TGB), a framework that significantly improves the temporal understanding and extrapolation capabilities of multimodal large language models for long-form video analysis.

Contribution

The paper presents a novel TGB framework with an efficient multi-span temporal grounding algorithm, a length extrapolation training paradigm, and a bootstrapping approach to enhance MLLMs without additional annotations.

Findings

TGB improves temporal grounding in MLLMs across seven benchmarks.

The model scales from 4 to 16 frames without performance loss.

Significant performance gains over prior MLLMs are demonstrated.

Abstract

Despite progress in multimodal large language models (MLLMs), the challenge of interpreting long-form videos in response to linguistic queries persists, largely due to the inefficiency in temporal grounding and limited pre-trained context window size. In this work, we introduce Temporal Grounding Bridge (TGB), a novel framework that bootstraps MLLMs with advanced temporal grounding capabilities and broadens their contextual scope. Our framework significantly enhances the temporal capabilities of current MLLMs through three key innovations: an efficient multi-span temporal grounding algorithm applied to low-dimension temporal features projected from flow; a multimodal length extrapolation training paradigm that utilizes low-dimension temporal features to extend the training context window size; and a bootstrapping framework that bridges our model with pluggable MLLMs without requiring annotation. We validate TGB across seven video benchmarks and demonstrate substantial performance improvements compared with prior MLLMs. Notably, our model, initially trained on sequences of four frames, effectively handles sequences up to 16 longer without sacrificing performance, highlighting its scalability and effectiveness in real-world applications. Our code is publicly available at https://github.com/bigai-nlco/VideoTGB