ResearchBenchmarksOff-policy evaluation on ScienceWorld (ETO)Off-policy evaluation on ScienceWorld (ETO)Follow0.97RhoWIS-0.03880.22310.4850.7469Jun 4, 2026Evaluation resultsMethodMethodLinksRhoCI95P-ValueWISn=5n=52026.060.97——ADWMn=5n=52026.060.8200.089FQEn=5n=52026.060.21——DRn=5n=52026.060.21——ISn=5n=52026.060.21——DMn=5n=52026.060——