PraxySante/qwen3-sft-asr-correction-v15-context
TEXT GENERATIONConcurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 4, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold
The PraxySante/qwen3-sft-asr-correction-v15-context model is a 0.8 billion parameter Qwen3-based language model. It is specifically fine-tuned for Automatic Speech Recognition (ASR) correction tasks, leveraging a substantial 32,768 token context window. This model is designed to improve the accuracy of transcribed speech by correcting errors, making it suitable for applications requiring high-fidelity text from audio.
Loading preview...
PraxySante/qwen3-sft-asr-correction-v15-context: ASR Correction Model
This model, developed by PraxySante, is a specialized 0.8 billion parameter variant based on the Qwen3 architecture. It is specifically fine-tuned for Automatic Speech Recognition (ASR) correction, aiming to enhance the accuracy of machine-generated transcripts.
Key Capabilities
- ASR Error Correction: Designed to identify and rectify inaccuracies in ASR outputs.
- Large Context Window: Features a significant 32,768 token context length, allowing it to process and correct longer segments of text with better contextual understanding.
- Optimized Performance: The model's training indicates consistent evaluation losses around 0.0732, suggesting stable and effective fine-tuning for its intended task.
Good For
- Improving ASR Accuracy: Ideal for post-processing ASR transcripts to reduce word error rates.
- Applications Requiring High-Fidelity Transcriptions: Suitable for use cases such as medical dictation, legal transcription, meeting minutes, and customer service call analysis where precise text is critical.
- Context-Aware Corrections: The extended context window makes it particularly effective for correcting errors that depend on broader conversational flow or sentence structure.