ResearchBenchmarksExecutable World-Model Learning on Baba in Wonderland extra-level transitions (held-out)Follow95.6Overall AccuracyAlice69.39276.1968389.804May 16, 2026Evaluation ResultsMethodMethodLinksOverall AccuracyBalanced AccuracyAliceEvaluation Protocol=Of...Evaluation Protocol=Offline, Calls=100, LLM Backbone=GPT-5.42026.0595.693.2CWMEvaluation Protocol=Of...Evaluation Protocol=Offline, Calls=100, LLM Backbone=GPT-5.42026.0582.673.4GIF-MCTSEvaluation Protocol=Of...Evaluation Protocol=Offline, Calls=100, LLM Backbone=GPT-5.42026.0570.462.8