Paper page - EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual Reasoning

Source: https://huggingface.co/papers/2610.28444 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

EvaOmni is an evidence-grounded benchmark that evaluates whether omnimodal language models identify audio-visual evidence and construct reasoning chains for audiovisual questions.

Omni LLMscan naturally locate key audio and visual evidence and integrate information across modalities foraudiovisual reasoning. Existing audio-visual reasoning benchmarks that mainly rely on multiple-choice questions usually score models solely based on answer correctness, making it difficult to assess whether a model truly forms a reasoning chain from evidence to the correct answer. In addition, textual or unimodal shortcuts may allow models to answer correctly without analyzing the necessary audio-visual evidence, further weakening the reliability of evaluation benchmarks. To address these issues, we introduce EvaOmni, anevidence-grounded benchmarkcontaining 392 videos and 1,027 purely human-annotated question-answer pairs, covering 15 capabilities and three levels: retrieval and localization, basic understanding, and complex reasoning. We design a series of rigorous annotation procedures in which experts annotate audio and visual evidence and construct video-based distractors; humans review whether questions meet strict inspection criteria, and finally models filter textual and unimodal shortcuts. Beyond high-quality question-answer pairs, we require models to simultaneously return the answer and supporting evidence intervals in order to evaluate the complete reasoning chain. We propose theCoverage Grounded Score(C.G.) to measure how much annotated evidence a model identifies, and theBottleneck Grounded Score(B.G.) to penalize poortemporal localizationin either modality; both are incorporated into answer correctness. Across 7 open-weight and 17 proprietaryOmni LLMs, Gemini3.7-Flash achieved the highest overall accuracy (79.4%), C.G. (70.8%), and B.G. (35.1%). At Level 3, it reached 79.1% accuracy, but its B.G. was only 12.3%. These results show that even the most advanced models still have room for development in grounded audio-visual reasoning.

View arXiv pageView PDFGitHub1Add to collection

Get this paper in your agent:

hf papers read 2610\.28444

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.28444 in a model README.md to link it from this page.

Datasets citing this paper1

#### VisionXLab/EvaOmni-Dataset Viewer• Updatedabout 4 hours ago • 1.03k

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.28444 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.