Entity · paper

Audio-Native Speech Recognition with a Frozen Discrete-Diffusion Language Model

paperactiveprovisionalaudio-native-speech-recognition-with-a-frozen-discrete-diffusion-language-model-156e4b0a·1 events·first seen 27h ago

Aliases: Audio-Native Speech Recognition with a Frozen Discrete-Diffusion Language Model

Co-occurring entities

DiffusionGemma LibriSpeech Google Connectionist Temporal Classification Whisper

More like this (12)

Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models Diffusion Language Models Self-Augmenting Retrieval for Diffusion Language Models LESS: Mutual-Stability Sampling for Diffusion Language Models DirectAudioEdit: Inversion-Free Text-Guided Audio Editing via Diffusion Prediction Contrast continuous diffusion language model Learning to Hear Hesitation: Continual Learning for Disfluency-Aware ASR Random Language Model Interleaved Speech Language Models Latently Work In Text Accelerating Masked Diffusion Large Language Models: A Survey of Efficient Inference Techniques FlowEdit: Associative Memory for Lifelong Pronunciation Adaptation in Flow-Matching TTS Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data

Recent events (1)

6arXiv · cs.AI·27h ago·source ↗

Discrete-diffusion language model adapted for parallel speech recognition via frozen DiffusionGemma

Researchers demonstrate that a frozen 26B mixture-of-experts discrete diffusion language model (DiffusionGemma) can be adapted for automatic speech recognition using only ~42M trainable parameters (0.16% of backbone). The system uses a frozen Whisper encoder for acoustic features, lightweight projectors, and LoRA adapters, achieving 6.6% WER on LibriSpeech test-clean in roughly eight parallel denoising steps regardless of utterance length. A key finding is that standard training objectives fail to ground audio features due to attention dismissal, and a CTC loss through the frozen output head resolves this. The approach supports multilingual transcription (English, Hindi, Mandarin) from a single adapter.

Frontier Model Releases Multimodal Progress DiffusionGemma LibriSpeech Google +3 more