LaSR: Context-Aware Speech Recognition via Latent Reasoning

Excerpt

arXiv:2606.00507v2 Announce Type: replace Abstract: Speech recognition in specialized domains requires leveraging contextual or topical information to improve the recognition of domain-specific entities. Speech Large Language Models (Speech LLMs) have substantially advanced speech understanding and reasoning capabilities, making context-aware speech recognition possible without predefined bias lists. In this paper, we propose LaSR (Latent Speech Reasoning), a novel training paradigm featuring a