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

TL;DR

This paper introduces LAMB, a novel framework that enhances audio captioning by aligning audio embeddings with LLM text space using Cauchy-Schwarz divergence, resulting in improved reasoning and state-of-the-art results.

Contribution

LAMB is the first to explicitly bridge the modality gap between audio and LLM embeddings with a divergence-based alignment and a two-stream adapter for richer audio features.

Findings

Achieves state-of-the-art performance on AudioCaps.

Effectively aligns audio and text embeddings at global and token levels.

Enhances reasoning capabilities of LLM-based audio captioning.

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.