SocialOmni: Benchmarking Audio-Visual Social Interactivity in Omni Models

TL;DR

SocialOmni introduces a new benchmark for evaluating audio-visual social interactivity in omni-modal large language models, focusing on dynamic conversational cues and robustness, revealing gaps between perception and interaction capabilities.

Contribution

This work presents SocialOmni, a comprehensive benchmark for assessing social interactivity in omni-modal models, including perception and generation tasks, with a detailed diagnostic set and analysis of current models.

Findings

Significant variance in social-interaction capabilities across models.

Perceptual accuracy does not correlate strongly with interaction quality.

Diagnostics provide actionable insights for improving social interactivity in models.

Abstract

Omni-modal large language models (OLMs) redefine human-machine interaction by natively integrating audio, vision, and text. However, existing OLM benchmarks remain anchored to static, accuracy-centric tasks, leaving a critical gap in assessing social interactivity, the fundamental capacity to navigate dynamic cues in natural dialogues. To this end, we propose SocialOmni, a comprehensive benchmark that operationalizes the evaluation of this conversational interactivity across three core dimensions: (i) speaker separation and identification (who is speaking), (ii) interruption timing control (when to interject), and (iii) natural interruption generation (how to phrase the interruption). SocialOmni features 2,000 perception samples and a quality-controlled diagnostic set of 209 interaction-generation instances with strict temporal and contextual constraints, complemented by controlled audio-visual inconsistency scenarios to test model robustness. We benchmarked 12 leading OLMs, which uncovers significant variance in their social-interaction capabilities across models. Furthermore, our analysis reveals a pronounced decoupling between a model's perceptual accuracy and its ability to generate contextually appropriate interruptions, indicating that understanding-centric metrics alone are insufficient to characterize conversational social competence. More encouragingly, these diagnostics from SocialOmni yield actionable signals for bridging the perception-interaction divide in future OLMs.