Incentivizing Temporal-Awareness in Egocentric Video Understanding Models

TL;DR

This paper introduces TGPO, a reinforcement learning algorithm that enhances temporal awareness in multimodal large language models for egocentric video understanding by explicitly rewarding temporal coherence.

Contribution

The paper proposes TGPO, a novel RLVR-based method that improves temporal reasoning in MLLMs, addressing their tendency to rely on spatial shortcuts.

Findings

TGPO improves temporal grounding accuracy across five egocentric video benchmarks.

TGPO outperforms prior RL-based approaches in causal coherence tasks.

TGPO effectively suppresses spatial shortcut behaviors in MLLMs.

Abstract

Multimodal large language models (MLLMs) have recently shown strong performance in visual understanding, yet they often lack temporal awareness, particularly in egocentric settings where reasoning depends on the correct ordering and evolution of events. This deficiency stems in part from training objectives that fail to explicitly reward temporal reasoning and instead rely on frame-level spatial shortcuts. To address this limitation, we propose Temporal Global Policy Optimization (TGPO), a reinforcement learning with verifiable rewards (RLVR) algorithm designed to incentivize temporal awareness in MLLMs. TGPO contrasts model outputs generated from temporally ordered versus shuffled video frames to derive calibrated, globally normalized reward signals that explicitly favor temporally coherent reasoning. Integrated with GRPO and GSPO, TGPO supports cold-start RL training and effectively suppresses spatial shortcut behaviors learned by existing MLLMs. Experiments across five egocentric video benchmarks demonstrate that TGPO consistently improves temporal grounding and causal coherence, outperforming prior RL-based video reasoning approaches. Our results suggest that TGPO offers a simple and scalable pathway toward temporally robust MLLMs for egocentric video understanding.