Do MLLMs Understand Pointing? Benchmarking and Enhancing Referential Reasoning in Egocentric Vision

TL;DR

This paper introduces EgoPoint-Bench, a new benchmark for evaluating and improving multimodal pointing reasoning in egocentric vision, addressing the limitations of current models in grounding spatial semantics of pointing gestures.

Contribution

It presents a comprehensive benchmark with synthetic and real data, demonstrating that fine-tuning models on synthetic data improves egocentric pointing performance and generalization.

Findings

State-of-the-art models struggle with egocentric pointing.

Fine-tuning on synthetic data improves performance and generalization.

The benchmark spans diverse evaluation dimensions and complexity levels.

Abstract

Egocentric AI agents, such as smart glasses, rely on pointing gestures to resolve referential ambiguities in natural language commands. However, despite advancements in Multimodal Large Language Models (MLLMs), current systems often fail to precisely ground the spatial semantics of pointing. Instead, they rely on spurious correlations with visual proximity or object saliency, a phenomenon we term "Referential Hallucination." To address this gap, we introduce EgoPoint-Bench, a comprehensive question-answering benchmark designed to evaluate and enhance multimodal pointing reasoning in egocentric views. Comprising over 11k high-fidelity simulated and real-world samples, the benchmark spans five evaluation dimensions and three levels of referential complexity. Extensive experiments demonstrate that while state-of-the-art proprietary and open-source models struggle with egocentric pointing, models fine-tuned on our synthetic data achieve significant performance gains and robust sim-to-real generalization. This work highlights the importance of spatially aware supervision and offers a scalable path toward precise egocentric AI assistants. Project page: https://guyyyug.github.io/EgoPoint-Bench/