Speech Question Answering on Trivia QA
39.1Speech QA ScoreGLM-4-Voice
Evaluation results
| Method | Links | |
|---|---|---|
| GLM-4-Voice2025.08 | 39.1 | |
| TurnGuideloss ratio=2:12025.08 | 32 | |
| Scaling InterleaveModel=9B, Train Params=9B, Data=600B tok2026.06 | 26.5 | |
| Moshi2025.08 | 22.8 | |
| Scaling InterleaveModel=9B, Train Params=9B, Data=200B tok2026.06 | 18.7 | |
| Ours (8B)Model=8B, Train Params=~150M, Data=2.5k hrs2026.06 | 17.6 | |
| SpeechGPT2025.08 | 14.8 | |
| Ours (4B)Model=4B, Train Params=~100M, Data=2.5k hrs2026.06 | 11.9 | |
| Scaling InterleaveModel=9B, Train Params=9B, Data=100B tok2026.06 | 11.7 | |
| MoshiModel=7B, Train Params=7B, Data=7M hrs2026.06 | 7.3 |