Kepler: Auditable World Models for ARC-AGI-3

arXiv:2610.00834 · cs.AI, cs.MA · Submitted 2026-09-30 · Read on arXiv

cs.AI, cs.MA

Submitted: 2026-09-30

Updated: 2026-09-30

Comments: 17 pages. Accepted to the non-archival Interpreting Agent Behavior workshop at NeurIPS 2026. Project: https://kepler-harness.vercel.app/ . Code and public traces available

Code: https://github.com/ryanbbrown/Retrodict

Project page: https://vista-research.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• World Model — 1× in this paper · explained in 7 episodes

Sources

• Tycho: Active Abstraction with Programmatic World Models for ARC-AGI-3
• When a Verified World Model Still Loses: Play-Adequacy vs Prediction-Accuracy in LLM-Synthesized Code World Models
• ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence
• On the Measure of Intelligence
• OPINE-World: Programmatic World Modeling with Ontology-error-Prioritized Interactive Exploration for ARC-AGI-3

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems