Phoneme-Guided Initialization for LLM-based Speech Recognition

eess.AS, cs.CL, cs.SD

Submitted: 2026-10-06

Updated: 2026-10-06

Comments: Accepted at IEEE SLT 2026

Code: https://github.com/zeerakahmed/makhzan

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• AudioPaLM: A Large Language Model That Can Speak and Listen
• Moshi: a speech-text foundation model for real-time dialogue
• Phonemes vs. Projectors: An Investigation of Speech-Language Interfaces for LLM-based ASR
• An Embarrassingly Simple Approach for LLM with Strong ASR Capacity
• Phoneme-Based Contextualization for Cross-Lingual Speech Recognition in End-to-End Models
• Phoneme-First Prediction for LLM-Based Speech Recognition
• Gaussian Error Linear Units (GELUs)

Related papers

• Rethinking Entropy Allocation in LLM-based ASR: Understanding the Dynamics between Speech Encoders and LLMs
• X-VC: Zero-shot Streaming Voice Conversion in Codec Space
• Autoregressive Guidance of Deep Spatially Selective Filters using Bayesian Tracking for Efficient Extraction of Moving Speakers
• VM-ARRAYDPS: Virtual Microphone Augmented Diffusion Posterior Sampling for Unsupervised Blind Speech Separation
• Conditional Flow Matching for Visually-Guided Acoustic Highlighting
• Anonymization, Not Elimination: Utility-Preserved Speech Anonymization