{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:5LYGAKK4WVHP2VRWXYHVIQNQWZ","short_pith_number":"pith:5LYGAKK4","schema_version":"1.0","canonical_sha256":"eaf060295cb54efd5636be0f5441b0b6491c36f4c01b7632628354b426e71009","source":{"kind":"arxiv","id":"2305.18354","version":2},"attestation_state":"computed","paper":{"title":"LLMs and the Abstraction and Reasoning Corpus: Successes, Failures, and the Importance of Object-based Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Elias B. Khalil, Pashootan Vaezipoor, Scott Sanner, Wenhao Li, Yudong Xu","submitted_at":"2023-05-26T16:32:17Z","abstract_excerpt":"Can a Large Language Model (LLM) solve simple abstract reasoning problems? We explore this broad question through a systematic analysis of GPT on the Abstraction and Reasoning Corpus (ARC), a representative benchmark of abstract reasoning ability from limited examples in which solutions require some \"core knowledge\" of concepts such as objects, goal states, counting, and basic geometry. GPT-4 solves only 13/50 of the most straightforward ARC tasks when using textual encodings for their two-dimensional input-output grids. Our failure analysis reveals that GPT-4's capacity to identify objects an"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2305.18354","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-26T16:32:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0e8692e7baeb96f4f654728b201be3600a34bae356a39a91a9facde03d116466","abstract_canon_sha256":"93a5385a9573cc7473d36d24103fd323074dd7f8d63d4737fca0d271f3725a14"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:45:14.017072Z","signature_b64":"at8WPqBiIwts9VSOpAxdVjBAwziAa1xTH4EKvGhKxALOu0SH9chbTpBpZh2i73Sjz0cM0/Z7luaFTWSpBxCZCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eaf060295cb54efd5636be0f5441b0b6491c36f4c01b7632628354b426e71009","last_reissued_at":"2026-07-05T07:45:14.015869Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:45:14.015869Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLMs and the Abstraction and Reasoning Corpus: Successes, Failures, and the Importance of Object-based Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Elias B. Khalil, Pashootan Vaezipoor, Scott Sanner, Wenhao Li, Yudong Xu","submitted_at":"2023-05-26T16:32:17Z","abstract_excerpt":"Can a Large Language Model (LLM) solve simple abstract reasoning problems? We explore this broad question through a systematic analysis of GPT on the Abstraction and Reasoning Corpus (ARC), a representative benchmark of abstract reasoning ability from limited examples in which solutions require some \"core knowledge\" of concepts such as objects, goal states, counting, and basic geometry. GPT-4 solves only 13/50 of the most straightforward ARC tasks when using textual encodings for their two-dimensional input-output grids. Our failure analysis reveals that GPT-4's capacity to identify objects an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.18354","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2305.18354/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2305.18354","created_at":"2026-07-05T07:45:14.016626+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.18354v2","created_at":"2026-07-05T07:45:14.016626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.18354","created_at":"2026-07-05T07:45:14.016626+00:00"},{"alias_kind":"pith_short_12","alias_value":"5LYGAKK4WVHP","created_at":"2026-07-05T07:45:14.016626+00:00"},{"alias_kind":"pith_short_16","alias_value":"5LYGAKK4WVHP2VRW","created_at":"2026-07-05T07:45:14.016626+00:00"},{"alias_kind":"pith_short_8","alias_value":"5LYGAKK4","created_at":"2026-07-05T07:45:14.016626+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.31031","citing_title":"GraphARC: A Comprehensive Benchmark for Graph-Based Abstract Reasoning","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2505.14412","citing_title":"PRL: Prompts from Reinforcement Learning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2510.08804","citing_title":"MOSAIC: Multi-agent Orchestration for Task-Intelligent Scientific Coding","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2512.11013","citing_title":"PIAST: Rapid Prompting with In-context Augmentation for Scarce Training data","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23106","citing_title":"No Test Cases, No Problem: Distillation-Driven Code Generation for Scientific Workflows","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12018","citing_title":"LLMs Struggle with Abstract Meaning Comprehension More Than Expected","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18839","citing_title":"One Step Forward and K Steps Back: Better Reasoning with Denoising Recursion Models","ref_index":168,"is_internal_anchor":false},{"citing_arxiv_id":"2402.01207","citing_title":"Efficient Causal Graph Discovery Using Large Language Models","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ","json":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ.json","graph_json":"https://pith.science/api/pith-number/5LYGAKK4WVHP2VRWXYHVIQNQWZ/graph.json","events_json":"https://pith.science/api/pith-number/5LYGAKK4WVHP2VRWXYHVIQNQWZ/events.json","paper":"https://pith.science/paper/5LYGAKK4"},"agent_actions":{"view_html":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ","download_json":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ.json","view_paper":"https://pith.science/paper/5LYGAKK4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.18354&json=true","fetch_graph":"https://pith.science/api/pith-number/5LYGAKK4WVHP2VRWXYHVIQNQWZ/graph.json","fetch_events":"https://pith.science/api/pith-number/5LYGAKK4WVHP2VRWXYHVIQNQWZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ/action/storage_attestation","attest_author":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ/action/author_attestation","sign_citation":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ/action/citation_signature","submit_replication":"https://pith.science/pith/5LYGAKK4WVHP2VRWXYHVIQNQWZ/action/replication_record"}},"created_at":"2026-07-05T07:45:14.016626+00:00","updated_at":"2026-07-05T07:45:14.016626+00:00"}