Tool Sequence Prediction on Multimedia
92.97Tool F1GRAFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GRAFTLLM=Llama-3.2-3B2026.05 | 92.97 | 0.11 | 90 | 76 | |
| GRAFTLLM=Mistral-7B2026.05 | 92.67 | 0.11 | 89 | 74 | |
| ToolGenLLM=Mistral-7B2026.05 | 91.18 | 0.14 | 88 | 72 | |
| GRAFTLLM=Qwen2.5-1.5B2026.05 | 90.52 | 0.16 | 85 | 66 | |
| ToolGenLLM=Llama-3.2-3B2026.05 | 89.45 | 0.15 | 86 | 69 | |
| ToolGenLLM=Qwen2.5-1.5B2026.05 | 88.36 | 0.18 | 84 | 62 | |
| GToolLLM=Mistral-7B2026.05 | 87.53 | 0.19 | 80 | 62 | |
| GToolLLM=Llama-3.2-3B2026.05 | 85.59 | 0.21 | 79 | 62 | |
| GToolLLM=Qwen2.5-1.5B2026.05 | 77.39 | 0.24 | 60 | 49 | |
| GNN4PlanLLM=Llama-3.2-3B2026.05 | 74.44 | 0.33 | 69 | 48 | |
| BeamSearchLLM=Mistral-7B2026.05 | 74 | 0.37 | 57 | 45 | |
| GNN4PlanLLM=Mistral-7B2026.05 | 72.63 | 0.36 | 68 | 44 | |
| GNN4PlanLLM=Qwen2.5-1.5B2026.05 | 68.13 | 0.42 | 59 | 36 | |
| DirectLLM=Mistral-7B2026.05 | 65.8 | 0.47 | 52 | 19 | |
| PLaGLLM=Llama-3.2-3B2026.05 | 56.58 | 0.52 | 46 | 14 | |
| HuggingGPTLLM=Llama-3.2-3B2026.05 | 55.62 | 0.55 | 44 | 13 | |
| PLaGLLM=Mistral-7B2026.05 | 55.62 | 0.55 | 40 | 13 | |
| DirectLLM=Llama-3.2-3B2026.05 | 54.51 | 0.58 | 44 | 15 | |
| HuggingGPTLLM=Mistral-7B2026.05 | 53.48 | 0.56 | 41 | 13 | |
| HuggingGPTLLM=Qwen2.5-1.5B2026.05 | 44.62 | 0.63 | 38 | 9 | |
| PLaGLLM=Qwen2.5-1.5B2026.05 | 43.24 | 0.66 | 35 | 7 | |
| DirectLLM=Qwen2.5-1.5B2026.05 | 42.99 | 0.65 | 37 | 7 | |
| BeamSearchLLM=Llama-3.2-3B2026.05 | 40.7 | 0.69 | 26 | 9 | |
| BeamSearchLLM=Qwen2.5-1.5B2026.05 | 37.72 | 0.71 | 22 | 6 |