Automatic Speech Recognition on SPGISpeech
1.84WERAudio Flamingo 3 + CoAT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Audio Flamingo 3 + CoATBackbone=Audio Flamingo 3, Method=CoAT2026.06 | 1.84 | — | — | |
| Audio Flamingo 3Backbone=Audio Flamingo 3, Method=Baseline2026.06 | 1.86 | — | — | |
| Qwen2.5-Omni-7B + CoATBackbone=Qwen2.5-Omni-7B, Method=CoAT2026.06 | 2.1 | — | — | |
| Qwen2.5-Omni-7BBackbone=Qwen2.5-Omni-7B, Method=Baseline2026.06 | 2.79 | — | — | |
| Qwen2-Audio + CoATBackbone=Qwen2-Audio, Method=CoAT2026.06 | 2.86 | — | — | |
| Whisper-large-v3-turbo# Total Params=1.5B2025.11 | 2.97 | — | — | |
| Qwen2.5-Omni-3B# Total Params=5B2025.11 | 3.24 | — | — | |
| whisper-large-v3Model Size=Large, Version=v32025.02 | 3.4 | 0.4 | — | |
| whisper-large-v3-turboModel Size=Large, Version=v3-turbo2025.02 | 3.4 | 0.4 | — | |
| whisper-mediumModel Size=Medium2025.02 | 3.7 | 0.2 | — | |
| whisper-largeModel Size=Large2025.02 | 3.7 | 0.6 | — | |
| distil-large-v2Model Size=Large, Type=Distilled, Version=v22025.02 | 4.1 | 1 | — | |
| whisper-small.enModel Size=Small, Language=English2025.02 | 4.1 | 0.9 | — | |
| LFM2-Audio-1.5B# Total Params=1.5B2025.11 | 4.17 | — | — | |
| whisper-medium.enModel Size=Medium, Language=English2025.02 | 4.3 | 1 | — | |
| whisper-large-v2Model Size=Large, Version=v22025.02 | 4.3 | 0.8 | — | |
| distil-medium.enModel Size=Medium, Type=Distilled, Language=English2025.02 | 4.6 | 0.6 | — | |
| distil-small.enModel Size=Small, Type=Distilled, Language=English2025.02 | 4.6 | 0.6 | — | |
| Qwen2-Audio-7BArchitecture=Qwen2, Parameters=7B2025.02 | 4.6 | 2.7 | — | |
| Qwen2-AudioBackbone=Qwen2-Audio, Method=Baseline2026.06 | 6.93 | — | — | |
| TE2SLSource=LibriSpeech2026.05 | 8.5 | — | 50.1 | |
| whisper-tinyModel Size=Tiny2025.02 | 8.8 | 3.6 | — | |
| Upsample-and-MaskSource=LibriSpeech2026.05 | 9.1 | — | 45.6 | |
| BaselineSource=LibriSpeech2026.05 | 11.1 | — | 39.4 | |
| Soft PromptSource=LibriSpeech2026.05 | 11.1 | — | 39.3 | |
| speechllm-1.5BParameters=1.5B2025.02 | 11.5 | 4.8 | — | |
| hubert-large-ls960-ftModel Size=Large, Finetuning=ls9602025.02 | 12.4 | 1.4 | — | |
| hf-seamless-m4t-mediumModel Size=Medium2025.02 | 13.2 | 4.4 | — | |
| hf-seamless-m4t-largeModel Size=Large2025.02 | 14.7 | 3.6 | — | |
| wav2vec2-large-xlsr-53-englishModel Size=Large, Architecture=wav2vec2-xlsr, Language=English2025.02 | 19.6 | 1 | — |