JAEGER: Joint 3D Audio-Visual Grounding and Reasoning in Simulated Physical Environments

TL;DR

JAEGER advances audio-visual large language models by integrating 3D spatial reasoning with RGB-D and multi-channel audio, enabling improved source localization and understanding in complex physical environments.

Contribution

It introduces a novel 3D extension for AV-LLMs, including a neural intensity vector for spatial audio and a large-scale benchmark for training and evaluation.

Findings

Outperforms 2D-based models in spatial reasoning tasks

Effectively estimates direction-of-arrival in noisy environments

Enhances understanding of complex physical scenes

Abstract

Current audio-visual large language models (AV-LLMs) are predominantly restricted to 2D perception, relying on RGB video and monaural audio. This design choice introduces a fundamental dimensionality mismatch that precludes reliable source localization and spatial reasoning in complex 3D environments. We address this limitation by presenting JAEGER, a framework that extends AV-LLMs to 3D space, to enable joint spatial grounding and reasoning through the integration of RGB-D observations and multi-channel first-order ambisonics. A core contribution of our work is the neural intensity vector (Neural IV), a learned spatial audio representation that encodes robust directional cues to enhance direction-of-arrival estimation, even in adverse acoustic scenarios with overlapping sources. To facilitate large-scale training and systematic evaluation, we propose SpatialSceneQA, a benchmark of 61k instruction-tuning samples curated from simulated physical environments. Extensive experiments demonstrate that our approach consistently surpasses 2D-centric baselines across diverse spatial perception and reasoning tasks, underscoring the necessity of explicit 3D modelling for advancing AI in physical environments. Our source code, pre-trained model checkpoints and datasets will be released upon acceptance.