Problem
Research questions and friction points this paper is trying to address.
Temporal Binding
Large Audio Language Models
Mechanistic Interpretability
Temporal Reasoning
Event Localization
On Temporal Binding in Large Audio Language Models
🤖 AI Summary
This study investigates the internal mechanisms by which large audio-language models (LALMs) process the temporal structure of sound events, which remain poorly understood. Employing mechanistic interpretability, representation analysis, and activation steering techniques, we examine how temporal information is bound to sound events within LALMs. Our findings reveal that event positional information is encoded along low-dimensional curved trajectories within intermediate-layer entity name representations. Intervening in these representations alters temporal reasoning beliefs without affecting precise timestamp predictions. This work demonstrates that coarse-grained temporal reasoning and fine-grained localization rely on independent neural pathways, thereby elucidating the internal temporal binding mechanisms of LALMs.
📝 Abstract
Reasoning about temporal structure of audio recordings requires Large Audio Language Models (LALMs) to associate sound events with their temporal position. Understanding the underlying mechanisms is a first step toward diagnosing failures and identifying model components that may need improvement. Using mechanistic interpretability, we investigate how temporal information is represented and bound to sound events in three open-source LALMs. We find that across all three, event-specific location becomes concentrated in event name representations at intermediate modality integration layers. These representations encode coarse event position along a low-dimensional, curved relative time trajectory. Steering event name representations along this trajectory systematically shifts before/after beliefs, providing evidence that these representations contribute to coarse temporal reasoning. In contrast, the same interventions do not reliably shift predicted onset timestamps, suggesting that coarse temporal reasoning and precise metric event localization rely on distinct mechanisms.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Mechanistic Interpretability
Large Audio Language Models
Temporal Binding
Representation Steering
Temporal Reasoning
🔎 Similar Papers
No similar papers found.
