Light-Omni: Reflex over Reasoning in Agentic Video Understanding with Long-Term Memory

2026-07-06

AI Summary

This work addresses the high latency and computational overhead of existing video agents, which rely on costly iterative inference to process long-horizon multimodal streams. To overcome these limitations, we propose Light-Omni, a novel framework that introduces a dual-context coupling mechanism combining global state and parameterized latent states, enabling iteration-free, low-latency video understanding and response within a single forward pass. Light-Omni achieves efficient long-term memory management and action control through hierarchical memory fusion, multimodal script construction, and parameterized latent state generation. Experimental results demonstrate that Light-Omni improves average accuracy by 2.4% across multiple video benchmarks, accelerates inference by 12.1×, and enhances GPU memory efficiency by 2.6×, while effectively boosting the performance of existing multimodal large language models.

Abstract

Agentic video understanding equips models with long-term memory to autonomously process and respond to continuous, long-horizon multimodal streams. However, advanced video agents often rely on ``detective-style'' iterative reasoning for action control (e.g., $\mathtt{search}$) and evidence aggregation, incurring prohibitive costs and latency. We argue that such heavy reasoning primarily compensates for the lack of global context and semantic misalignment in retrieval. This paper introduces Light-Omni, a multimodal agent framework for reflexive and lightweight video understanding. It achieves this through dual contextual states that instantly build the required context in a single forward pass. First, we maintain a global state, a finite-sized multimodal script continuously consolidated from episodic memory, serving as the global context for Light-Omni. Through hierarchical merging, it preserves recent details while summarizing past events. Second, conditioned on this global context, we generate a parametric latent state that directly drives autonomous actions and produces retrieval embeddings, with minimal latency. Benefiting from this coupled design, Light-Omni achieves semantically aligned retrieval and reflexive responses while avoiding iterative reasoning. Extensive experiments validate the effectiveness of Light-Omni across multiple video benchmarks. Notably, it outperforms M3-Agent with an average 2.4% accuracy gain, a 12.1$\times$ speedup, and a 2.6$\times$ improvement in GPU memory efficiency. Furthermore, it serves as a memory system to enhance both the performance and efficiency of existing MLLMs. Project page: https://clare-nie.github.io/Light-Omni.

Problem

• agentic video understanding
• long-term memory
• iterative reasoning
• multimodal streams
• latency

Innovation

• reflexive video understanding
• long-term memory
• multimodal agent
• global context
• latent state