# Transcribe, Translate, and Optimize: Joint Reward Learning for Speech Translation

arXiv: https://arxiv.org/abs/2609.26536 · HTML: https://mlsanity.com/paper/2609.26536/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning.
- Builds on 22 of its 30 references that are in the map.
- First posted on arXiv 22 Sep 2026.

## Placed in

- LLM reasoning / [Reinforcement learning for reasoning](https://mlsanity.com/lr/reinforcement-learning-for-reasoning/)

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

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [Qwen2.5-Omni Technical Report](https://mlsanity.com/paper/2503.20215/) (2025)
- [FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech](https://mlsanity.com/paper/2205.12446/) (2022)
- [SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuning](https://mlsanity.com/paper/2408.05517/) (2024)
- [Sequence-to-Sequence Models Can Directly Translate Foreign Speech](https://mlsanity.com/paper/1703.08581/) (2017)
- [Listen and Translate: A Proof of Concept for End-to-End Speech-to-Text Translation](https://mlsanity.com/paper/1612.01744/) (2016)
- [Minimum Word Error Rate Training for Attention-based Sequence-to-Sequence Models](https://mlsanity.com/paper/1712.01818/) (2017)
- [Tied Multitask Learning for Neural Speech Translation](https://mlsanity.com/paper/1802.06655/) (2018)
- [Speech Translation with Large Language Models: An Industrial Practice](https://mlsanity.com/paper/2312.13585/) (2023)
- [Making LLMs Better Many-to-Many Speech-to-Text Translators with Curriculum Learning](https://mlsanity.com/paper/2409.19510/) (2024)
- [Chain-of-Thought Prompting for Speech Translation](https://mlsanity.com/paper/2409.11538/) (2024)
- [Improving Rare Word Recognition with LM-aware MWER Training](https://mlsanity.com/paper/2204.07553/) (2022)
- [LLaST: Improved End-to-end Speech Translation System Leveraged by Large Language Models](https://mlsanity.com/paper/2407.15415/) (2024)
- [Group Relative Policy Optimization for Speech Recognition](https://mlsanity.com/paper/2509.01939/) (2025)
- [Advancing Speech Understanding in Speech-Aware Language Models with GRPO](https://mlsanity.com/paper/2509.16990/) (2025)
- [Speech-to-Text Translation with Phoneme-Augmented CoT: Enhancing Cross-Lingual Transfer in Low-Resource Scenarios](https://mlsanity.com/paper/2505.24691/) (2025)
- [Revisiting Direct Speech-to-Text Translation with Speech LLMs: Better Scaling than CoT Prompting?](https://mlsanity.com/paper/2510.03093/) (2025)
- [End-to-end Automatic Speech Recognition and Speech Translation: Integration of Speech Foundational Models and LLMs](https://mlsanity.com/paper/2510.10329/) (2025)
- [Listening or Reading? Evaluating Speech Awareness in Chain-of-Thought Speech-to-Text Translation](https://mlsanity.com/paper/2510.03115/) (2025)

## Abstract (arXiv)

In LLM-based speech translation, transcription-based chain-of-thought (CoT) suffers from a mismatch between reference transcripts used in supervised fine-tuning (SFT) and model-generated transcripts at inference. To address this, we propose joint recognition and translation fine-tuning via group relative policy optimization (GRPO). We score both transcripts and translations, with translation conditioned on model-generated transcripts, and compare three token advantage strategies. Using Qwen2.5-Omni-3B across four languages, we evaluate CoT against direct speech translation (Direct ST) under SFT and GRPO, training on CoVoST 2 and testing on CoVoST 2 and FLEURS. CoT GRPO outperforms Direct ST GRPO by 1.77 and 0.83 average BLEU points on CoVoST 2 and FLEURS. Compared to CoT SFT, GRPO boosts BLEU by 0.82 and 0.67 points and reduces word error rate (WER) by 8.8% and 7.2% relatively. These results highlight reinforcement fine-tuning as an effective method to mitigate the training-inference mismatch, jointly improving recognition and translation.