# RRPO: Robust Reward Policy Optimization for LLM-based Emotional TTS

arXiv: https://arxiv.org/abs/2512.04552 · HTML: https://mlsanity.com/paper/2512.04552/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Reward modeling and RLHF, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 22 of its 24 references that are in the map.
- First posted on arXiv 4 Dec 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward modeling and RLHF](https://mlsanity.com/alignment/reward-modeling-and-rlhf/)

## Builds on (most cited first, 22 of 22)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [mixup: Beyond Empirical Risk Minimization](https://mlsanity.com/paper/1710.09412/) (2017)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://mlsanity.com/paper/2110.13900/) (2021)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Seed-TTS: A Family of High-Quality Versatile Speech Generation Models](https://mlsanity.com/paper/2406.02430/) (2024)
- [CosyVoice 2: Scalable Streaming Speech Synthesis with Large Language Models](https://mlsanity.com/paper/2412.10117/) (2024)
- [Adversarial Training Methods for Semi-Supervised Text Classification](https://mlsanity.com/paper/1605.07725/) (2016)
- [Emotional Voice Conversion: Theory, Databases and ESD](https://mlsanity.com/paper/2105.14762/) (2021)
- [InstructTTS: Modelling Expressive TTS in Discrete Latent Space with Natural Language Style Prompt](https://mlsanity.com/paper/2301.13662/) (2023)
- [EmoDiff: Intensity Controllable Emotional Text-to-Speech with Soft-Label Guidance](https://mlsanity.com/paper/2211.09496/) (2022)
- [EmoVoice: LLM-based Emotional Text-To-Speech Model with Freestyle Text Prompting](https://mlsanity.com/paper/2504.12867/) (2025)
- [MER 2023: Multi-label Learning, Modality Robustness, and Semi-Supervised Learning](https://mlsanity.com/paper/2304.08981/) (2023)
- [Emo-DPO: Controllable Emotional Speech Synthesis through Direct Preference Optimization](https://mlsanity.com/paper/2409.10157/) (2024)
- [VoxInstruct: Expressive Human Instruction-to-Speech Generation with Unified Multilingual Codec Language Modelling](https://mlsanity.com/paper/2408.15676/) (2024)
- [Preference Alignment Improves Language Model-Based TTS](https://mlsanity.com/paper/2409.12403/) (2024)
- [EmoMix: Emotion Mixing via Diffusion Models for Emotional Speech Synthesis](https://mlsanity.com/paper/2306.00648/) (2023)
- [Differentiable Reward Optimization for LLM based TTS system](https://mlsanity.com/paper/2507.05911/) (2025)
- [ZET-Speech: Zero-shot adaptive Emotion-controllable Text-to-Speech Synthesis with Diffusion and Style-based Models](https://mlsanity.com/paper/2305.13831/) (2023)

## Built on by (most cited first, 4 of 4)

- [FlowTTS-GRPO: Online Reinforcement Learning with Multi-Objective Reward Optimization for Flow-Matching Based Text-to-Speech](https://mlsanity.com/paper/2606.23190/) (2026)
- [WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training](https://mlsanity.com/paper/2604.14932/) (2026)
- [HPRO: Hierarchical Progressive Reward Optimization via Preference Extraction for Emotional Text-to-Speech](https://mlsanity.com/paper/2606.28249/) (2026)
- [On the Emotion Understanding of Synthesized Speech](https://mlsanity.com/paper/2603.16483/) (2026)

## Abstract (arXiv)

Differentiable reinforcement learning (RL) frameworks like DiffRO offer a powerful approach for controllable text-to-speech (TTS), but are vulnerable to reward hacking, particularly for nuanced tasks like emotion control. The policy model can exploit a vanilla Reward Model (RM) by generating acoustic artifacts to achieve spurious rewards, but at the cost of degrading perceptual quality. To address this, we propose Robust Reward Policy Optimization (RRPO), a novel framework that employs a hybrid regularization scheme. This scheme develops a robust RM whose reward signal is more reliably aligned with human perception, compelling the policy to abandon detrimental shortcuts and instead learn the complex features of genuine emotions. Our ablation study confirms the enhanced robustness of our RM, as evidenced by its strong cross-lingual generalization. The subjective evaluation demonstrates that this robust RM effectively mitigates reward hacking, leading to significant improvements in both emotional expressiveness and naturalness over all baselines. Demo page: https://lrwinr.github.io/RRPO-CosyVoice.