arXiv cs.CL· Tim Wenjie Hu, Evelyn Nafula Ouma, Benjamin Akera, John A. Quinn·· 13 小时前AI 评分31
用 GRPO 提升低资源语言语音识别：Whisper-large-v3 微调出 SunflowerASR
Group Relative Policy Optimisation Improves Multilingual Speech Recognition in Low-Resource Languages
AI 导读
研究者将 Group Relative Policy Optimisation（GRPO）用于非 LLM 的小型编码器-解码器 ASR 模型，以 Whisper-large-v3 为案例覆盖 51 种非洲语言，训练数据稀缺或音频含噪时收益最明显。
正文 · 原文
Abstract:Autoregressive encoder-decoder speech recognition models trained with teacher forcing can suffer from exposure bias, limiting their ability to recover from their own prediction errors at inference time. Recent work has shown that reinforcement learning with Group Relative Policy Optimisation (GRPO) improves speech recognition in major languages with LLM-based models. We extend this line of work to smaller, non-LLM encoder-decoder ASR models, using Whisper-large-v3 as a case study across 51 African languages. The benefits of this approach are most pronounced when training data is scarce or when input audio is noisy or degraded. Our resulting model, SunflowerASR, achieves the best published macro-averaged WER on the AfriVox-v2 (0.227) and SimbaBench (0.372) benchmarks, and is publicly available.
| Comments: | 5 pages, 3 figures, submitted to ICASSP 2027 |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.30663 [cs.AI] |
| (or arXiv:2610.30663v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.30663 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tim Wenjie Hu [view email]
[v1]
Thu, 22 Oct 2026 03:24:38 UTC (959 KB)
来源：arXiv cs.CL · arxiv.org