RLBR: Reinforcement Learning with Biasing Rewards for Contextual Speech Large Language Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915739716288512 |
|---|---|
| author | Ren, Bo Fan, Ruchao Shen, Yelong Chen, Weizhu Li, Jinyu |
| author_facet | Ren, Bo Fan, Ruchao Shen, Yelong Chen, Weizhu Li, Jinyu |
| contents | Speech large language models (LLMs) have driven significant progress in end-to-end speech understanding and recognition, yet they continue to struggle with accurately recognizing rare words and domain-specific terminology. This paper presents a novel fine-tuning method, Reinforcement Learning with Biasing Rewards (RLBR), which employs a specialized biasing words preferred reward to explicitly emphasize biasing words in the reward calculation. In addition, we introduce reference-aware mechanisms that extend the reinforcement learning algorithm with reference transcription to strengthen the potential trajectory exploration space. Experiments on the LibriSpeech corpus across various biasing list sizes demonstrate that RLBR delivers substantial performance improvements over a strong supervised fine-tuning (SFT) baseline and consistently outperforms several recently published methods. The proposed approach achieves excellent performance on the LibriSpeech test-clean and test-other sets, reaching Biasing Word Error Rates (BWERs) of 0.59% / 2.11%, 1.09% / 3.24%, and 1.36% / 4.04% for biasing list sizes of 100, 500, and 1000, respectively, without compromising the overall WERs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_13409 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | RLBR: Reinforcement Learning with Biasing Rewards for Contextual Speech Large Language Models Ren, Bo Fan, Ruchao Shen, Yelong Chen, Weizhu Li, Jinyu Audio and Speech Processing Speech large language models (LLMs) have driven significant progress in end-to-end speech understanding and recognition, yet they continue to struggle with accurately recognizing rare words and domain-specific terminology. This paper presents a novel fine-tuning method, Reinforcement Learning with Biasing Rewards (RLBR), which employs a specialized biasing words preferred reward to explicitly emphasize biasing words in the reward calculation. In addition, we introduce reference-aware mechanisms that extend the reinforcement learning algorithm with reference transcription to strengthen the potential trajectory exploration space. Experiments on the LibriSpeech corpus across various biasing list sizes demonstrate that RLBR delivers substantial performance improvements over a strong supervised fine-tuning (SFT) baseline and consistently outperforms several recently published methods. The proposed approach achieves excellent performance on the LibriSpeech test-clean and test-other sets, reaching Biasing Word Error Rates (BWERs) of 0.59% / 2.11%, 1.09% / 3.24%, and 1.36% / 4.04% for biasing list sizes of 100, 500, and 1000, respectively, without compromising the overall WERs. |
| title | RLBR: Reinforcement Learning with Biasing Rewards for Contextual Speech Large Language Models |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2601.13409 |