# CARD: Cross-component Audio Representation Distillation for Encoder-Free Audio Captioning

arXiv: https://arxiv.org/abs/2607.04619 · HTML: https://mlsanity.com/paper/2607.04619/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 31 references that are in the map.
- First posted on arXiv 6 Jul 2026.

## Builds on (most cited first, 25 of 25)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [CIDEr: Consensus-based Image Description Evaluation](https://mlsanity.com/paper/1411.5726/) (2014)
- [SPICE: Semantic Propositional Image Caption Evaluation](https://mlsanity.com/paper/1607.08822/) (2016)
- [Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation](https://mlsanity.com/paper/2211.06687/) (2022)
- [Clotho: An Audio Captioning Dataset](https://mlsanity.com/paper/1910.09387/) (2019)
- [WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research](https://mlsanity.com/paper/2303.17395/) (2023)
- [HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detection](https://mlsanity.com/paper/2202.00874/) (2022)
- [MiniLLM: On-Policy Distillation of Large Language Models](https://mlsanity.com/paper/2306.08543/) (2023)
- [Improved Image Captioning via Policy Gradient optimization of SPIDEr](https://mlsanity.com/paper/1612.00370/) (2016)
- [Unveiling Encoder-Free Vision-Language Models](https://mlsanity.com/paper/2406.11832/) (2024)
- [Clotho-AQA: A Crowdsourced Dataset for Audio Question Answering](https://mlsanity.com/paper/2204.09634/) (2022)
- [Mono-InternVL: Pushing the Boundaries of Monolithic Multimodal Large Language Models with Endogenous Visual Pre-training](https://mlsanity.com/paper/2410.08202/) (2024)
- [Infinity Instruct: Scaling Instruction Selection and Synthesis to Enhance Language Models](https://mlsanity.com/paper/2506.11116/) (2025)
- [Vision as LoRA](https://mlsanity.com/paper/2503.20680/) (2025)
- [Auto-ACD: A Large-scale Dataset for Audio-Language Representation Learning](https://mlsanity.com/paper/2309.11500/) (2023)
- [CED: Consistent ensemble distillation for audio tagging](https://mlsanity.com/paper/2308.11957/) (2023)
- [EnCLAP: Combining Neural Audio Codec and Audio-Text Joint Embedding for Automated Audio Captioning](https://mlsanity.com/paper/2401.17690/) (2024)
- [SLAM-AAC: Enhancing Audio Captioning with Paraphrasing Augmentation and CLAP-Refine through LLMs](https://mlsanity.com/paper/2410.09503/) (2024)
- [Enhancing Automated Audio Captioning via Large Language Models with Optimized Audio Encoding](https://mlsanity.com/paper/2406.13275/) (2024)
- [DRCap: Decoding CLAP Latents with Retrieval-Augmented Generation for Zero-shot Audio Captioning](https://mlsanity.com/paper/2410.09472/) (2024)
- [LLM can Read Spectrogram: Encoder-free Speech-Language Modeling](https://mlsanity.com/paper/2606.10231/) (2026)
- [AuRA: Internalizing Audio Understanding into LLMs as LoRA](https://mlsanity.com/paper/2606.11033/) (2026)

## Abstract (arXiv)

Modern automated audio captioning systems pair a frozen audio encoder with a large language model (LLM) via a trainable projector, incurring the encoder's inference cost and bottlenecking the model through its fixed acoustic features. We present CARD, an encoder-free audio captioning model that removes the encoder at inference: a 13.2M projector feeds a frozen LLM with merged LoRA adapters, while the teacher used to train it is discarded. CARD distills a pre-trained audio teacher (CLAP-HTSAT) into the model, but rather than injecting it into the LLM alone, it routes the teacher's representations across components: perceptual stages to the projector and semantic stages to the LLM. This placement improves CIDEr-D by +11.9 over an LLM-only distilled model on AudioCaps and by +5.0 on Clotho, reaching 55.4 against a 66.4 encoder-kept upper bound with no encoder at inference, showing that where a teacher's knowledge is placed matters as much as its presence.