Tool Learning on RestBench TMDB
86.2Success RateEVOTOOL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EVOTOOLBackbone=GPT-4o2026.03 | 86.2 | — | — | |
| ToolCoderBackbone=gpt-4o-mini, Paradigm=Code-based2025.02 | 85 | 78 | 83 | |
| DRAFTBackbone=GPT-4o2026.03 | 84.3 | — | — | |
| RestGPTBackbone=gpt-4o-mini, Paradigm=Text-based2025.02 | 83 | 33 | 49 | |
| ToolCoder w/o Reusable RepositoryBackbone=gpt-4o-mini, Paradigm=Code-based, Reusable Repository=false2025.02 | 83 | 71 | 78 | |
| EASYTOOLBackbone=GPT-4o2026.03 | 81.4 | — | — | |
| CodeActBackbone=gpt-4o-mini, Paradigm=Code-based2025.02 | 80 | 56 | 67 | |
| HarnessForgeBackbone=Qwen3-8B2026.06 | 80 | — | 88.75 | |
| ATCBackbone=gpt-4o-mini, Paradigm=Code-based2025.02 | 78 | 58 | 62 | |
| EvoPromptBackbone=GPT-4o2026.03 | 76.3 | — | — | |
| ReActBackbone=gpt-4o-mini, Paradigm=Text-based2025.02 | 76 | 48 | 50 | |
| HarnessForgeBackbone=Qwen3-4B2026.06 | 76 | — | 85.1 | |
| EVOTOOLBackbone=Qwen3-8B2026.03 | 75.9 | — | — | |
| ANYTOOLBackbone=GPT-4o2026.03 | 75.2 | — | — | |
| ChameleonBackbone=gpt-4o-mini, Paradigm=Text-based2025.02 | 75 | 45 | 52 | |
| EasyToolBackbone=gpt-4o-mini, Paradigm=Text-based2025.02 | 75 | 45 | 62 | |
| ToolCoder w/o Error ReflectionBackbone=gpt-4o-mini, Paradigm=Code-based, Error Reflection=false2025.02 | 75 | 65 | 77 | |
| DRAFTBackbone=Qwen3-8B2026.03 | 74.2 | — | — | |
| PromptBreederBackbone=GPT-4o2026.03 | 74.1 | — | — | |
| RLOOBackbone=Qwen3-8B2026.06 | 74 | — | 84.25 | |
| OPROBackbone=GPT-4o2026.03 | 73.8 | — | — | |
| ReActBackbone=GPT-4o2026.03 | 72.4 | — | — | |
| ConAgentsBackbone=gpt-4o-mini, Paradigm=Text-based2025.02 | 72 | 38 | 57 | |
| EASYTOOLBackbone=Qwen3-8B2026.03 | 71.6 | — | — | |
| GRPOBackbone=Qwen3-8B2026.06 | 70 | — | 83.2 | |
| SFTBackbone=Qwen3-8B2026.06 | 69 | — | 82.92 | |
| EvoPromptBackbone=Qwen3-8B2026.03 | 67.1 | — | — | |
| AdaPlannerBackbone=GPT-4o2026.03 | 66.2 | — | — | |
| ANYTOOLBackbone=Qwen3-8B2026.03 | 66.2 | — | — | |
| PromptBreederBackbone=Qwen3-8B2026.03 | 65.2 | — | — | |
| OPROBackbone=Qwen3-8B2026.03 | 64.9 | — | — | |
| Plan-and-SolveBackbone=GPT-4o2026.03 | 64.8 | — | — | |
| GRPOBackbone=Qwen3-4B2026.06 | 64 | — | 76.67 | |
| ReActBackbone=Qwen3-8B2026.03 | 63.7 | — | — | |
| SFTBackbone=Qwen3-4B2026.06 | 61 | — | 75.7 | |
| RLOOBackbone=Qwen3-4B2026.06 | 61 | — | 77.42 | |
| CoTBackbone=GPT-4o2026.03 | 59.8 | — | — | |
| AdaPlannerBackbone=Qwen3-8B2026.03 | 58.3 | — | — | |
| Plan-and-SolveBackbone=Qwen3-8B2026.03 | 57 | — | — | |
| CoTBackbone=Qwen3-8B2026.03 | 52.6 | — | — | |
| ADASBackbone=Qwen3-8B2026.06 | 51 | — | 60.41 | |
| MermaidFlowBackbone=Qwen3-8B2026.06 | 47 | — | 58.67 | |
| ADASBackbone=Qwen3-4B2026.06 | 45 | — | 57.43 | |
| MaASBackbone=Qwen3-8B2026.06 | 43 | — | 54.78 | |
| MermaidFlowBackbone=Qwen3-4B2026.06 | 39 | — | 51.95 | |
| AFlowBackbone=Qwen3-8B2026.06 | 39 | — | 51.28 | |
| AgentSquareBackbone=Qwen3-8B2026.06 | 38 | — | 53.83 | |
| MaASBackbone=Qwen3-4B2026.06 | 37 | — | 46.57 | |
| AgentSquareBackbone=Qwen3-4B2026.06 | 35 | — | 49.83 | |
| AFlowBackbone=Qwen3-4B2026.06 | 32 | — | 47.64 |