Listening with the Eyes: Benchmarking Egocentric Co-Speech Grounding across Space and Time

TL;DR

This paper introduces EcoG-Bench, a challenging benchmark for evaluating egocentric co-speech grounding that requires models to jointly predict what, where, and when, revealing significant gaps in current multimodal understanding.

Contribution

The paper presents EcoG-Bench, a new bilingual benchmark with dense annotations for evaluating egocentric speech-gesture grounding, and demonstrates the limitations of state-of-the-art models in this task.

Findings

Humans achieve near-ceiling performance (96.9%) on EcoG-Bench.

Current models perform poorly, with the best at 17.0%.

Improving audio-visual interface timing significantly boosts model performance.

Abstract

In situated collaboration, speakers often use intentionally underspecified deictic commands (e.g., ``pass me \textit{that}''), whose referent becomes identifiable only by aligning speech with a brief co-speech pointing \emph{stroke}. However, many embodied benchmarks admit language-only shortcuts, allowing MLLMs to perform well without learning the \emph{audio--visual alignment} required by deictic interaction. To bridge this gap, we introduce \textbf{Egocentric Co-Speech Grounding (EcoG)}, where grounding is executable only if an agent jointly predicts \textit{What}, \textit{Where}, and \textit{When}. To operationalize this, we present \textbf{EcoG-Bench}, an evaluation-only bilingual (EN/ZH) diagnostic benchmark of \textbf{811} egocentric clips with dense spatial annotations and millisecond-level stroke supervision. It is organized under a \textbf{Progressive Cognitive Evaluation} protocol. Benchmarking state-of-the-art MLLMs reveals a severe executability gap: while human subjects achieve near-ceiling performance on EcoG-Bench (\textbf{96.9\%} strict Eco-Accuracy), the best native video-audio setting remains low (Gemini-3-Pro: \textbf{17.0\%}). Moreover, in a diagnostic ablation, replacing the native video--audio interface with timestamped frame samples and externally verified ASR (with word-level timing) substantially improves the same model (\textbf{17.0\%}$\to$\textbf{42.9\%}). Overall, EcoG-Bench provides a strict, executable testbed for event-level speech--gesture binding, and suggests that multimodal interfaces may bottleneck the observability of temporal alignment cues, independently of model reasoning.