LAMB: LLM-based Audio Captioning with Modality Gap Bridging via Cauchy-Schwarz Divergence

TL;DR

LAMB addresses the modality gap between audio embeddings and LLM text embeddings in automated audio captioning by introducing a Cross-Modal Aligner that uses Cauchy–Schwarz divergence and mutual information, a Two-Stream Adapter to enrich audio representations, and a Token Guide to steer generation within the LLM embedding space. The framework jointly optimizes global and token-level alignment and decoding guidance, yielding state-of-the-art results on AudioCaps and competitive performance on Clotho without heavy external pretraining. This approach demonstrates that explicit cross-modal alignment in the LLM embedding space improves the reasoning capabilities of the decoder for audio understanding, enabling more faithful and coherent captions. The combination of CMA, TSA, and Token Guide provides a practical pathway to leverage powerful LLMs for AAC with improved fidelity and controllability.

Abstract

Automated Audio Captioning aims to describe the semantic content of input audio. Recent works have employed large language models (LLMs) as a text decoder to leverage their reasoning capabilities. However, prior approaches that project audio features into the LLM embedding space without considering cross-modal alignment fail to fully utilize these capabilities. To address this, we propose LAMB, an LLM-based audio captioning framework that bridges the modality gap between audio embeddings and the LLM text embedding space. LAMB incorporates a Cross-Modal Aligner that minimizes Cauchy-Schwarz divergence while maximizing mutual information, yielding tighter alignment between audio and text at both global and token levels. We further design a Two-Stream Adapter that extracts semantically enriched audio embeddings, thereby delivering richer information to the Cross-Modal Aligner. Finally, leveraging the aligned audio embeddings, a proposed Token Guide directly computes scores within the LLM text embedding space to steer the output logits of generated captions. Experimental results confirm that our framework strengthens the reasoning capabilities of the LLM decoder, achieving state-of-the-art performance on AudioCaps.

Figures (2)

• Figure 1: (a) Overview of the proposed LAMB framework with (b) Cross-Modal Aligner and (c) Two-Stream Adapter.
• Figure 2: UMAP Visualization of audio–text embedding alignment with/without Cross-Modal Aligner (CMA).