Beyond Prompting: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration (LOGIC)
This addresses the problem of scalable and robust entity recognition in speech models for users dealing with dynamic data like contact names or technical terms, offering an incremental improvement over existing methods.
The paper tackles the challenge of Speech LLMs recognizing domain-specific entities by introducing LOGIC, a framework that integrates contextual biasing directly in the decoding layer, achieving a 9% relative reduction in Entity WER with only a 0.30% increase in False Alarm Rate.
The rapid emergence of new entities -- driven by cultural shifts, evolving trends, and personalized user data -- poses a significant challenge for existing Speech Large Language Models (Speech LLMs). While these models excel at general conversational tasks, their static training knowledge limits their ability to recognize domain-specific terms such as contact names, playlists, or technical jargon. Existing solutions primarily rely on prompting, which suffers from poor scalability: as the entity list grows, prompting encounters context window limitations, increased inference latency, and the "lost-in-the-middle" phenomenon. An alternative approach, Generative Error Correction (GEC), attempts to rewrite transcripts via post-processing but frequently suffers from "over-correction", introducing hallucinations of entities that were never spoken. In this work, we introduce LOGIC (Logit-Space Integration for Contextual Biasing), an efficient and robust framework that operates directly in the decoding layer. Unlike prompting, LOGIC decouples context injection from input processing, ensuring constant-time complexity relative to prompt length. Extensive experiments using the Phi-4-MM model across 11 multilingual locales demonstrate that LOGIC achieves an average 9% relative reduction in Entity WER with a negligible 0.30% increase in False Alarm Rate.