Interactive Scientific Exploration on ScienceWorld standard 30-task protocol
73.72Average Score (Short)SDP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SDPLLM=GPT-4, Training status=Training-free2026.05 | 73.72 | 53.5 | 50.41 | 59.16 | |
| ReflexionLLM=GPT-4, Training status=Training-free2026.05 | 71.47 | 35.43 | 30.17 | 45.34 | |
| Plan-and-ActLLM=GPT-4, Training status=Training-free2026.05 | 60.52 | 46.43 | 34.77 | 47.86 | |
| CoTLLM=GPT-4, Training status=Training-free2026.05 | 49.54 | 47.87 | 23.09 | 39.23 | |
| ReActLLM=GPT-4, Training status=Training-free2026.05 | 48.79 | 44.01 | 21.07 | 36.43 | |
| EVOAGENTLLM=GPT-4, Training status=Training-free2026.05 | 48.67 | 36.17 | 11.38 | 30.42 | |
| SayCanLLM=GPT-4, Training status=Training-free2026.05 | 43.83 | 36.58 | 23.65 | 33.82 |