{"paper":{"title":"Executable World Models for ARC-AGI-3 in the Era of Coding Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A coding-agent system using executable Python world models solves 7 of 25 public ARC-AGI-3 games with no game-specific code.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Sergey Rodionov","submitted_at":"2026-05-06T17:12:36Z","abstract_excerpt":"We evaluate an initial coding-agent system for ARC-AGI-3 in which the agent maintains an executable Python world model, verifies it against previous observations, refactors it toward simpler abstractions as a practical proxy for an MDL-like simplicity bias, and plans through the model before acting. The system is intentionally direct: it uses a scripted controller, predefined world-model interfaces, verifier programs, and a plan executor, but no hand-coded game-specific logic. The agent-facing prompts, workspace, and controller contain no game-specific code, game-specific prompts, hand-coded h"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The agent fully solved 7 games, achieved a Relative Human Action Efficiency greater than 75% on 6 games, and obtained a mean per-game RHAE of 32.58%. Because the system uses no game-specific code, it can serve as a game-general baseline for ARC-AGI-3.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That verifier-driven executable Python world models with refactoring as a proxy for MDL-like simplicity bias will generalize to the private validation set and other unseen tasks without requiring game-specific logic.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Executable world model agents solve 7 of 25 ARC-AGI-3 public games with 32.58% mean RHAE as a game-general baseline.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A coding-agent system using executable Python world models solves 7 of 25 public ARC-AGI-3 games with no game-specific code.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"426f977406652b2afb7c45b63ad52800054bc592b9032df3c33ba6337a8ee8de"},"source":{"id":"2605.05138","kind":"arxiv","version":2},"verdict":{"id":"31eb2c2b-de04-4a80-8766-f4325fc7de51","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T17:47:48.925371Z","strongest_claim":"The agent fully solved 7 games, achieved a Relative Human Action Efficiency greater than 75% on 6 games, and obtained a mean per-game RHAE of 32.58%. Because the system uses no game-specific code, it can serve as a game-general baseline for ARC-AGI-3.","one_line_summary":"Executable world model agents solve 7 of 25 ARC-AGI-3 public games with 32.58% mean RHAE as a game-general baseline.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That verifier-driven executable Python world models with refactoring as a proxy for MDL-like simplicity bias will generalize to the private validation set and other unseen tasks without requiring game-specific logic.","pith_extraction_headline":"A coding-agent system using executable Python world models solves 7 of 25 public ARC-AGI-3 games with no game-specific code."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05138/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:36:11.066307Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T21:31:19.495118Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T13:49:11.916293Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"3d1d61b9a383e85d57f48259e5bf87e31c3bdb46e8dca951986c81180fb60fb2"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a53259f1c6684c514591e6d48aada392657918e04854f0d9e3a5e60677618326"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}