Alice 系统如何在线自监督发现可执行世界模型：Baba in Wonderland 研究

View PDF HTML (experimental)

Abstract:Executable world models can be read, edited, executed, and reused for planning, but only if the program captures the environment's transition law rather than semantic shortcuts in its surface vocabulary. We study online executable world-model learning under prior misalignment, where an agent must induce state-dependent dynamics from interaction evidence alone, without rule descriptions, reward signals, or trustworthy lexical priors. We introduce Alice, a closed-loop system that treats failed candidate updates as structural signal: when a candidate explains a new transition but loses previously explained ones, the preservation conflict reveals dynamics that the current program had conflated. Alice refines these conflicts into hypothesis classes that both provide a few broken examples, one per affected class, for update and guide frontier exploration toward transitions that are novel and underrepresented with respect to the current program. We evaluate Alice on Baba in Wonderland, a prior-misaligned variant of Baba Is You that preserves simulator dynamics while replacing semantically meaningful rule-property labels with unrelated words. Experiments show that Alice substantially improves executable world-model learning under prior misalignment, and five-run ablations support overall improvement trends as class structure and exploration signals are added. Project page: this https URL.

Submission history

From: HyeongYeop Kang [view email]
[v1] Sat, 16 May 2026 00:18:22 UTC (2,298 KB)
[v2] Wed, 4 Nov 2026 11:01:26 UTC (3,783 KB)