Spoken Question Answering on LibriSQA (En)
69.5AccuracyFastSLM
Evaluation results
| Method | Links | |
|---|---|---|
| FastSLMModel Size=4.8B, #speech tokens/30 Sec.=502026.01 | 69.5 | |
| Gemini-2.5-Flash#speech tokens/30 Sec.=9602026.01 | 67 | |
| Phi4-MMModel Size=5.8B, #speech tokens/30 Sec.=3752026.01 | 64.5 | |
| Qwen2-AudioModel Size=8B, #speech tokens/30 Sec.=1012026.01 | 57.2 | |
| Voxtral-miniModel Size=4.7B, #speech tokens/30 Sec.=7502026.01 | 48.9 |