Unverified50% confidenceFactExact time
The two stages of PTC-Bias share the same set of phoneme posterior probabilities and do not require additional SpeechLLM forward passes.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPTC-Bias is a two-stage context biasing framework based on phoneme-level temporal competition, designed to improve the recognition of rare words in speech large language models.75% similarUnverifiedHMM假设语音波形由音素状态序列依概率生成，解码时通过维特比算法寻找后验概率最大的状态路径，并通常与高斯混合模型结合形成GMM-HMM架构61% similarUnverifiedAcoustic model inference is the primary source of TTS latency.61% similarUnverifiedEach time a TTS (Text-to-Speech) speech synthesis system is called, the model's random sampling may cause subtle drift in timbre, manifesting as inconsistencies in details such as speaking rate, intonation, and resonance characteristics.61% similarUnverified自回归模型（如VALL-E）逐token预测语音离散表征，能建模长程依赖和丰富韵律，但存在累积误差；非自回归模型（如FastSpeech）并行生成速度快但表现力受限61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949144API
curl https://kongchang.com/api/v1/knowledge/claims/949144MCP
get_claim(id=949144)