A Benchmark and Agentic Framework for Omni-Modal Reasoning and Tool Use in Long Videos

TL;DR

This paper introduces LongShOTBench, a comprehensive benchmark for evaluating multimodal reasoning and tool use in long videos, and presents LongShOTAgent, an agentic system that improves understanding of complex video content.

Contribution

It provides a new diagnostic benchmark with open-ended questions and a scalable, human-verified pipeline, along with an agentic system for long video analysis, addressing gaps in existing evaluation methods.

Findings

State-of-the-art models perform significantly below human-level on LongShOTBench.

Open-source models achieve below 30% accuracy, highlighting the challenge.

LongShOTAgent achieves 44.66% accuracy, demonstrating the potential of agentic approaches.

Abstract

Long-form multimodal video understanding requires integrating vision, speech, and ambient audio with coherent long-range reasoning. Existing benchmarks emphasize either temporal length or multimodal richness, but rarely both and while some incorporate open-ended questions and advanced metrics, they mostly rely on single-score accuracy, obscuring failure modes. We introduce LongShOTBench, a diagnostic benchmark with open-ended, intent-driven questions; single- and multi-turn dialogues; and tasks requiring multimodal reasoning and agentic tool use across video, audio, and speech. Each item includes a reference answer and graded rubric for interpretable, and traceable evaluation. LongShOTBench is produced via a scalable, human-validated pipeline to ensure coverage and reproducibility. All samples in our LongShOTBench are human-verified and corrected. Furthermore, we present LongShOTAgent, an agentic system that analyzes long videos via preprocessing, search, and iterative refinement. On LongShOTBench, state-of-the-art MLLMs show large gaps: Gemini-2.5-Flash achieves 52.95%, open-source models remain below 30%, and LongShOTAgent attains 44.66%. These results underscore the difficulty of real-world long-form video understanding. LongShOTBench provides a practical, reproducible foundation for evaluating and improving MLLMs. All resources are available on GitHub: https://github.com/mbzuai-oryx/longshot.