Scientific Interaction on ScienceWorld 1400 (test)
75.9Success RateCVT-RL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CVT-RLDescription=per-seed CVT-RL success, Split (train/dev/test)=9600/1200/1400, Individual Seeds=75.0/76.8/75.7/76.4/75.62026.06 | 75.9 | 3.2 | |
| IMDescription=information-matched counterfactual-process RL, Split (train/dev/test)=9600/1200/14002026.06 | 72.7 | — | |
| CMDescription=compute-matched non-causal RL, Split (train/dev/test)=9600/1200/14002026.06 | 69.1 | — |