{"version":1,"type":"story","url":"https://digestai.news/story/ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti","json":"https://digestai.news/story/ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti.json","markdown":"https://digestai.news/story/ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti.md","slug":"ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti","headline":"PTC-Bias improves speech LLM accuracy with phoneme-level bias correction","summary":"Researchers introduced **PTC-Bias**, a two-stage method for reducing errors in speech large language models (SpeechLLMs). The framework uses phoneme-level temporal competition to refine rare-word recognition. In the first stage, **PTC Retrieval** decodes phonemes frame-by-frame and narrows down candidate pronunciations into a shortlist. The second stage, **PTC Correction**, compares retrieved candidates against decoded text spans to correct near-homophone and segmentation errors without rerunning the SpeechLLM.\n\nTests on the **LibriSpeech** dataset show **PTC-Bias** cuts bias-word error rates by up to **23.4%** (test-clean) and **23.9%** (test-other) compared to **CTC-Filter**, while keeping unweighted error rates stable. The method works with bias lists of up to **2000 words** and two SpeechLLMs: **Prompt-SLAM-ASR-7B** and an unspecified second model. No extra SpeechLLM forward passes are needed, as both stages reuse phoneme posteriors.","keyPoints":["PTC-Bias uses phoneme-level temporal competition to filter and correct rare-word errors in SpeechLLMs","Tests on LibriSpeech show up to 23.9% relative reduction in bias-word error rates for 2000-word lists","Method requires no additional SpeechLLM forward passes, relying only on phoneme posteriors"],"whyItMatters":"This technique could lower misheard-word errors in voice assistants and transcription tools, especially for rare or domain-specific terms. It avoids costly model reruns by leveraging phoneme-level competition during decoding.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":[],"models":["Prompt-SLAM-ASR-7B"],"people":[]},"firstPublishedAt":"2026-09-25T04:00:00Z","updatedAt":"2026-09-25T04:00:00Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"arXiv cs.CL","title":"PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs","url":"https://arxiv.org/abs/2609.28727","publishedAt":"2026-09-25T04:00:00Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"PTC-Bias improves speech LLM accuracy with phoneme-level bias correction\", 25 September 2026, https://digestai.news/story/ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti","publisher":"Digest AI","title":"PTC-Bias improves speech LLM accuracy with phoneme-level bias correction","datePublished":"2026-09-25T04:00:00Z","url":"https://digestai.news/story/ptc-bias-improves-speech-llm-accuracy-with-phoneme-level-bias-correcti"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}