DRCap: Decoding CLAP Latents with Retrieval-Augmented Generation for Zero-shot Audio Captioning

TL;DR

DRCap tackles the data- and domain-robustness limitations of traditional automated audio captioning by uniting CLAP latent alignment with an LLM in a zero-shot setting. It trains on text-only data to reconstruct captions from CLAP text embeddings using a lightweight mapper, and performs inference by replacing the text encoder with the audio encoder, augmented by projection-based decoding and retrieval-augmented generation to bridge modality gaps. The approach introduces a similarity-selection scheme for robust training, a fixed prompting strategy, and domain-adaptive text embedding support with a large caption datastore to enable rapid domain transfer without retraining. Empirically, DRCap achieves strong in-domain performance on zero-shot AAC benchmarks and state-of-the-art cross-domain transfer, illustrating the practical impact of modular, data-efficient multimodal decoding for audio captioning.

Abstract

While automated audio captioning (AAC) has made notable progress, traditional fully supervised AAC models still face two critical challenges: the need for expensive audio-text pair data for training and performance degradation when transferring across domains. To overcome these limitations, we present DRCap, a data-efficient and flexible zero-shot audio captioning system that requires text-only data for training and can quickly adapt to new domains without additional fine-tuning. DRCap integrates a contrastive language-audio pre-training (CLAP) model and a large-language model (LLM) as its backbone. During training, the model predicts the ground-truth caption with a fixed text encoder from CLAP, whereas, during inference, the text encoder is replaced with the audio encoder to generate captions for audio clips in a zero-shot manner. To mitigate the modality gap of the CLAP model, we use both the projection strategy from the encoder side and the retrieval-augmented generation strategy from the decoder side. Specifically, audio embeddings are first projected onto a text embedding support to absorb extensive semantic information within the joint multi-modal space of CLAP. At the same time, similar captions retrieved from a datastore are fed as prompts to instruct the LLM, incorporating external knowledge to take full advantage of its strong generative capability. Conditioned on both the projected CLAP embedding and the retrieved similar captions, the model is able to produce a more accurate and semantically rich textual description. By tailoring the text embedding support and the caption datastore to the target domain, DRCap acquires a robust ability to adapt to new domains in a training-free manner. Experimental results demonstrate that DRCap outperforms all other zero-shot models in in-domain scenarios and achieves state-of-the-art performance in cross-domain scenarios.

Figures (1)

• Figure 1: Left: Overview of the CLAP model and the modality gap within its latent space. Right: Overview of the proposed DRCap. Based on the aligned multi-modal space of CLAP wu2023large, during training, DRCap learns to decode the text embedding to reconstruct the original caption. Only the linear mapping network $m$ is trained, while the LLM is fine-tuned using the LoRA hu2021lora method. The similarity selection was employed to prevent learning collapse caused by text-to-text retrieval. During inference, the audio embedding is first projected onto a text embedding support $\mathcal{S}$, mitigating the modality gap. The top-$k$ most similar captions retrieved from the datastore $\mathcal{DS}$ are used as prompts to instruct the LLM, producing accurate and semantically rich descriptions. Both the text embedding support $\mathcal{S}$ and the datastore $\mathcal{DS}$ could be changed in the inference stage, offering DRCap the flexibility to adapt to new domains.