Voice Interaction Evaluation on TPI-Bench Janus (test)
86RSFTPI-Base
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TPI-BaseData C (Trained on corpus data)=Yes, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 86 | 3.54 | |
| TPI-VAData C (Trained on corpus data)=Yes, Data VA (Trained on dyadic voice assistant data)=Yes, Data HN (Trained on hard negatives)=No2026.04 | 67 | 3.75 | |
| TPI-FullData C (Trained on corpus data)=Yes, Data VA (Trained on dyadic voice assistant data)=Yes, Data HN (Trained on hard negatives)=Yes2026.04 | 16 | 4.8 | |
| Kimi-Audio-Instruct-7BData C (Trained on corpus data)=No, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 13 | 4.52 | |
| Qwen2.5-Omni-7BData C (Trained on corpus data)=No, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 12 | 4.44 | |
| VITA-Audio-Instruct-7BData C (Trained on corpus data)=No, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 10 | 4.37 | |
| ChatGPT-4o-audioData C (Trained on corpus data)=No, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 9 | 4.95 | |
| Qwen3-Omni-30B-A3B-InstructData C (Trained on corpus data)=No, Data VA (Trained on dyadic voice assistant data)=No, Data HN (Trained on hard negatives)=No2026.04 | 1 | 4.45 |