# LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement

arXiv: https://arxiv.org/abs/2603.13952 · HTML: https://mlsanity.com/paper/2603.13952/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 41 references that are in the map.
- First posted on arXiv 14 Mar 2026.

## Builds on (most cited first, 26 of 26)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [SALMONN: Towards Generic Hearing Abilities for Large Language Models](https://mlsanity.com/paper/2310.13289/) (2023)
- [SDR - half-baked or well done?](https://mlsanity.com/paper/1811.02508/) (2018)
- [Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation](https://mlsanity.com/paper/1809.07454/) (2018)
- [LRS3-TED: a large-scale dataset for visual speech recognition](https://mlsanity.com/paper/1809.00496/) (2018)
- [DNSMOS: A Non-Intrusive Perceptual Objective Speech Quality metric to evaluate Noise Suppressors](https://mlsanity.com/paper/2010.15258/) (2020)
- [Supervised Speech Separation Based on Deep Learning: An Overview](https://mlsanity.com/paper/1708.07524/) (2017)
- [NISQA: A Deep CNN-Self-Attention Model for Multidimensional Speech Quality Prediction with Crowdsourced Datasets](https://mlsanity.com/paper/2104.09494/) (2021)
- [MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement](https://mlsanity.com/paper/2104.03538/) (2021)
- [MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement](https://mlsanity.com/paper/1905.04874/) (2019)
- [An Overview of Deep-Learning-Based Audio-Visual Speech Enhancement and Separation](https://mlsanity.com/paper/2008.09586/) (2020)
- [CMGAN: Conformer-based Metric GAN for Speech Enhancement](https://mlsanity.com/paper/2203.15149/) (2022)
- [SpeechBERTScore: Reference-Aware Automatic Evaluation of Speech Generation Leveraging NLP Evaluation Metrics](https://mlsanity.com/paper/2401.16812/) (2024)
- [VERSA: A Versatile Evaluation Toolkit for Speech, Audio, and Music](https://mlsanity.com/paper/2412.17667/) (2024)
- [Audio-visual Speech Enhancement Using Conditional Variational Auto-Encoders](https://mlsanity.com/paper/1908.02590/) (2019)
- [ESPnet-Codec: Comprehensive Training and Evaluation of Neural Codecs for Audio, Music, and Speech](https://mlsanity.com/paper/2409.15897/) (2024)
- [Audio Large Language Models Can Be Descriptive Speech Quality Evaluators](https://mlsanity.com/paper/2501.17202/) (2025)
- [QualiSpeech: A Speech Quality Assessment Dataset with Natural Language Reasoning and Descriptions](https://mlsanity.com/paper/2503.20290/) (2025)
- [Reinforcement Learning Based Speech Enhancement for Robust Speech Recognition](https://mlsanity.com/paper/1811.04224/) (2018)
- [Self-Supervised Speech Quality Estimation and Enhancement Using Only Clean Speech](https://mlsanity.com/paper/2402.16321/) (2024)
- [A Study on Zero-shot Non-intrusive Speech Assessment using Large Language Models](https://mlsanity.com/paper/2409.09914/) (2024)
- [Using RLHF to align speech enhancement approaches to mean-opinion quality scores](https://mlsanity.com/paper/2410.13182/) (2024)
- [AV-CrossNet: an Audiovisual Complex Spectral Mapping Network for Speech Separation By Leveraging Narrow- and Cross-Band Modeling](https://mlsanity.com/paper/2406.11619/) (2024)

## Abstract (arXiv)

In existing Audio-Visual Speech Enhancement (AVSE) methods, objectives such as Scale-Invariant Signal-to-Noise Ratio (SI-SNR) and Mean Squared Error (MSE) are widely used; however, their correlation with perceived speech quality is often suboptimal and provides limited interpretability for optimization. This work proposes a reinforcement learning-based AVSE framework with a Large Language Model (LLM)-based interpretable reward model. An audio LLM generates natural language descriptions of enhanced speech, which are converted by a sentiment analysis model into a 1-5 rating score serving as the PPO reward for fine-tuning a pretrained AVSE model. Compared with scalar metrics, LLM-generated feedback is semantically rich and explicitly describes speech quality improvements. Experiments on the AVSEC-4 dataset show that the proposed method outperforms a supervised baseline and a DNSMOS-based RL baseline in PESQ, STOI, neural quality metrics, and subjective listening tests.