Function Selection on TaskBench Multimedia
82.3Function Selection AccRaw Demonstrations
Evaluation Results
| Method | Links | |
|---|---|---|
| Raw DemonstrationsBackbone=Qwen3 14B2025.12 | 82.3 | |
| Raw DemonstrationsBackbone=GPT-4o2025.12 | 82.3 | |
| No ICLBackbone=Qwen3 14B2025.12 | 81.7 | |
| No ICLBackbone=GPT-4o2025.12 | 81.6 | |
| Raw DemonstrationsBackbone=Qwen3 8B2025.12 | 80.4 | |
| No ICLBackbone=Qwen3 8B2025.12 | 78.5 | |
| DTDR-CBackbone=GPT-4o2025.12 | 74.8 | |
| DTDR-CBackbone=Qwen3 14B2025.12 | 74 | |
| Dependency RetrievalBackbone=Qwen3 14B2025.12 | 73.8 | |
| DTDR-LBackbone=GPT-4o2025.12 | 73.7 | |
| DTDR-LBackbone=Qwen3 14B2025.12 | 73.4 | |
| Dependency RetrievalBackbone=GPT-4o2025.12 | 72.8 | |
| Dependency RetrievalBackbone=Qwen3 8B2025.12 | 69.1 | |
| DTDR-CBackbone=Qwen3 8B2025.12 | 69.1 | |
| DTDR-LBackbone=Qwen3 8B2025.12 | 67.6 | |
| DTDR-LBackbone=Qwen3 4B2025.12 | 64.1 | |
| Learned RetrievalBackbone=Qwen3 4B2025.12 | 58.2 | |
| Raw DemonstrationsBackbone=Qwen3 4B2025.12 | 57.3 | |
| No ICLBackbone=Qwen3 4B2025.12 | 54.3 | |
| DTDR-CBackbone=Qwen3 4B2025.12 | 52.4 | |
| Dependency RetrievalBackbone=Qwen3 4B2025.12 | 51.8 | |
| Learned RetrievalBackbone=GPT-4o2025.12 | 51.1 | |
| Learned RetrievalBackbone=Qwen3 14B2025.12 | 50.8 | |
| Learned RetrievalBackbone=Qwen3 8B2025.12 | 49.6 | |
| DTDR-CBackbone=Qwen3 1.7B2025.12 | 47.8 | |
| Raw DemonstrationsBackbone=Qwen3 1.7B2025.12 | 46.9 | |
| DTDR-LBackbone=Qwen3 1.7B2025.12 | 45.5 | |
| Dependency RetrievalBackbone=Qwen3 1.7B2025.12 | 45.1 | |
| Learned RetrievalBackbone=Qwen3 1.7B2025.12 | 42.1 | |
| No ICLBackbone=Qwen3 1.7B2025.12 | 38.1 | |
| DTDR-CBackbone=Qwen3 0.6B2025.12 | 20.2 | |
| DTDR-LBackbone=Qwen3 0.6B2025.12 | 17.6 | |
| Learned RetrievalBackbone=Qwen3 0.6B2025.12 | 12.1 | |
| Dependency RetrievalBackbone=Qwen3 0.6B2025.12 | 9.5 | |
| No ICLBackbone=Qwen3 0.6B2025.12 | 7.3 | |
| Raw DemonstrationsBackbone=Qwen3 0.6B2025.12 | 5.3 |