Function Selection on TaskBench DailyLife
96.8Function Selection AccuracyRaw Demonstrations
Evaluation Results
| Method | Links | |
|---|---|---|
| Raw DemonstrationsBackbone=GPT-4o2025.12 | 96.8 | |
| Raw DemonstrationsBackbone=Qwen3 14B2025.12 | 96.6 | |
| No ICLBackbone=GPT-4o2025.12 | 96.6 | |
| No ICLBackbone=Qwen3 14B2025.12 | 96.2 | |
| Raw DemonstrationsBackbone=Qwen3 8B2025.12 | 94.5 | |
| No ICLBackbone=Qwen3 8B2025.12 | 93.9 | |
| DTDR-LBackbone=Qwen3 14B2025.12 | 93.7 | |
| DTDR-LBackbone=GPT-4o2025.12 | 91.5 | |
| DTDR-LBackbone=Qwen3 8B2025.12 | 91 | |
| DTDR-LBackbone=Qwen3 4B2025.12 | 89 | |
| Raw DemonstrationsBackbone=Qwen3 4B2025.12 | 87.6 | |
| Dependency RetrievalBackbone=Qwen3 14B2025.12 | 84.1 | |
| DTDR-CBackbone=Qwen3 14B2025.12 | 84.1 | |
| No ICLBackbone=Qwen3 4B2025.12 | 83.5 | |
| DTDR-LBackbone=Qwen3 1.7B2025.12 | 83.1 | |
| Dependency RetrievalBackbone=GPT-4o2025.12 | 82.9 | |
| DTDR-CBackbone=GPT-4o2025.12 | 82.9 | |
| Dependency RetrievalBackbone=Qwen3 8B2025.12 | 75.3 | |
| DTDR-CBackbone=Qwen3 8B2025.12 | 75.3 | |
| Raw DemonstrationsBackbone=Qwen3 1.7B2025.12 | 69 | |
| No ICLBackbone=Qwen3 1.7B2025.12 | 66.7 | |
| Dependency RetrievalBackbone=Qwen3 4B2025.12 | 64.7 | |
| DTDR-CBackbone=Qwen3 4B2025.12 | 64.7 | |
| Learned RetrievalBackbone=Qwen3 4B2025.12 | 60.4 | |
| DTDR-CBackbone=Qwen3 1.7B2025.12 | 60.2 | |
| Dependency RetrievalBackbone=Qwen3 1.7B2025.12 | 60 | |
| Learned RetrievalBackbone=Qwen3 1.7B2025.12 | 56.6 | |
| Learned RetrievalBackbone=Qwen3 14B2025.12 | 56.5 | |
| Learned RetrievalBackbone=Qwen3 8B2025.12 | 52.2 | |
| Learned RetrievalBackbone=GPT-4o2025.12 | 48.1 | |
| DTDR-LBackbone=Qwen3 0.6B2025.12 | 45.9 | |
| DTDR-CBackbone=Qwen3 0.6B2025.12 | 23.7 | |
| Learned RetrievalBackbone=Qwen3 0.6B2025.12 | 19.6 | |
| No ICLBackbone=Qwen3 0.6B2025.12 | 12.8 | |
| Dependency RetrievalBackbone=Qwen3 0.6B2025.12 | 11.5 | |
| Raw DemonstrationsBackbone=Qwen3 0.6B2025.12 | 10.5 |