Emotional Support Dialogue on ESConv (SR/AT)
52.5Success Rate (SR)w. UKA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| w. UKABackbone=Seed-1.6-36B2026.05 | 52.5 | 7.28 | |
| w. UKABackbone=Qwen3-235B-A22B2026.05 | 48.3 | 7.46 | |
| w. PRINCIPLESBackbone=Qwen3-235B-A22B2026.05 | 45.2 | 7.51 | |
| w. PRINCIPLESBackbone=Seed-1.6-36B2026.05 | 45 | 7.41 | |
| Prompting BaselineBackbone=Qwen3-235B-A22B2026.05 | 43.6 | 7.58 | |
| w. UKABackbone=GPT-OSS-120B2026.05 | 40.1 | 7.56 | |
| w. MetaMindBackbone=GPT-OSS-120B2026.05 | 39.6 | 7.59 | |
| w. PRINCIPLESBackbone=GPT-OSS-120B2026.05 | 38.8 | 7.76 | |
| Prompting BaselineBackbone=GPT-OSS-120B2026.05 | 38.2 | 7.6 | |
| w. MetaMindBackbone=Qwen3-235B-A22B2026.05 | 38.2 | 8.02 | |
| w. UKABackbone=Qwen3-32B2026.05 | 35.4 | 7.81 | |
| w. MetaMindBackbone=Qwen3-32B2026.05 | 33.8 | 7.97 | |
| w. PRINCIPLESBackbone=Qwen3-32B2026.05 | 33 | 8.07 | |
| Prompting BaselineBackbone=Qwen3-32B2026.05 | 30.6 | 7.93 | |
| w. MetaMindBackbone=Seed-1.6-36B2026.05 | 24.3 | 7.82 | |
| Prompting BaselineBackbone=Seed-1.6-36B2026.05 | 21.5 | 8.03 |