Omni-DeepSearch: A Benchmark for Audio-Driven Omni-Modal Deep Search

TL;DR

This paper introduces Omni-DeepSearch, a new benchmark for evaluating models that perform audio-driven omni-modal deep search involving multi-hop reasoning across multiple modalities.

Contribution

It presents a novel benchmark with a multi-stage filtering pipeline to evaluate the ability of models to perform audio-based cross-modal search and reasoning.

Findings

The strongest model achieves only 43.44% accuracy, indicating high task difficulty.

Key bottlenecks include audio entity inference, query formulation, and multi-hop retrieval.

The benchmark reveals significant challenges in audio-driven omni-modal reasoning.

Abstract

Current omni-modal benchmarks mainly evaluate models under settings where multiple modalities are provided simultaneously, while the ability to start from audio alone and actively search for cross-modal evidence remains underexplored. In this paper, we introduce \textbf{Omni-DeepSearch}, a benchmark for audio-driven omni-modal deep search. Given one or more audio clips and a related question, models must infer useful clues from audio, invoke text, image, and video search tools, and perform multi-hop reasoning to produce a short, objective, and verifiable answer. Omni-DeepSearch contains 640 samples across 15 fine-grained categories, covering four retrieval target modalities and four audio content types. A multi-stage filtering pipeline ensures audio dependence, retrieval necessity, visual modality necessity, and answer uniqueness. Experiments on recent closed-source and open-source omni-modal models show that this task remains highly challenging: the strongest evaluated model, Gemini-3-Pro, achieves only 43.44\% average accuracy. Further analyses illustrate key bottlenecks in audio entity inference, query formulation, tool-use reliability, multi-hop retrieval, and cross-modal verification. These results highlight audio-driven omni-modal deep search as an important and underexplored direction for future multimodal agents.