Task Completion on ScienceWorld Seen
10.2Average StepsEAGLET
Evaluation Results
| Method | Links | |
|---|---|---|
| EAGLETExecutor Agent=GPT-52025.10 | 10.2 | |
| w/o GuidanceExecutor Agent=GPT-52025.10 | 11.3 | |
| MPOExecutor Agent=GPT-52025.10 | 12.1 | |
| EAGLETExecutor Agent=GPT-4.12025.10 | 12.2 | |
| MPOExecutor Agent=GPT-4.12025.10 | 13.6 | |
| w/o GuidanceExecutor Agent=GPT-4.12025.10 | 14.4 |