ResearchBenchmarksTeapot-Mug cross-category generalization on SimulationFollow55Success RateAffordGen-1.47213.18927.8542.511Apr 12, 2026Evaluation ResultsMethodMethodLinksSuccess RateAffordGenZero-shot=trueZero-shot=true2026.0455CPGenZero-shot=trueZero-shot=true2026.042.7DemoGenZero-shot=trueZero-shot=true2026.040.7