{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VJIBQPNWAF4QR34T6QK4RUDCRK","short_pith_number":"pith:VJIBQPNW","schema_version":"1.0","canonical_sha256":"aa50183db6017908ef93f415c8d0628aa972677d06f74df98203127d4bed3d55","source":{"kind":"arxiv","id":"2505.07859","version":2},"attestation_state":"computed","paper":{"title":"Product of Experts with LLMs: Boosting Performance on ARC Is a Matter of Perspective","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Daniel Franzen, David Hartmann, Jan Disselhoff","submitted_at":"2025-05-08T11:17:10Z","abstract_excerpt":"The Abstraction and Reasoning Corpus (ARC-AGI) poses a significant challenge for large language models (LLMs), exposing limitations in their abstract reasoning abilities. In this work, we leverage task-specific data augmentations throughout the training, generation, and scoring phases, and employ a depth-first search algorithm to generate diverse, high-probability candidate solutions. Furthermore, we utilize the LLM not only as a generator but also as a scorer, using its output probabilities to select the most promising solutions. Our method achieves a score of 71.6% (286.5/400 solved tasks) o"},"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":"2505.07859","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-08T11:17:10Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"37c33287736d5655c2e5a560714dd60daf22d00b854a62ca2c0250400bf47e2e","abstract_canon_sha256":"a60829258dae52e105396780b2ae4fab4aa76a192e4b569f19aee68194351d8b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:19:52.605208Z","signature_b64":"rMBNaVGTguHj/6/ToiZx4ydKQ4L7B6m6GRuAQkd6XZloS9ijGiXvgndrutIJqyJxZrBml+80hdMq8KjgJVKIDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aa50183db6017908ef93f415c8d0628aa972677d06f74df98203127d4bed3d55","last_reissued_at":"2026-07-05T11:19:52.604765Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:19:52.604765Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Product of Experts with LLMs: Boosting Performance on ARC Is a Matter of Perspective","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Daniel Franzen, David Hartmann, Jan Disselhoff","submitted_at":"2025-05-08T11:17:10Z","abstract_excerpt":"The Abstraction and Reasoning Corpus (ARC-AGI) poses a significant challenge for large language models (LLMs), exposing limitations in their abstract reasoning abilities. In this work, we leverage task-specific data augmentations throughout the training, generation, and scoring phases, and employ a depth-first search algorithm to generate diverse, high-probability candidate solutions. Furthermore, we utilize the LLM not only as a generator but also as a scorer, using its output probabilities to select the most promising solutions. Our method achieves a score of 71.6% (286.5/400 solved tasks) o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.07859","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/2505.07859/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":"2505.07859","created_at":"2026-07-05T11:19:52.604824+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.07859v2","created_at":"2026-07-05T11:19:52.604824+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.07859","created_at":"2026-07-05T11:19:52.604824+00:00"},{"alias_kind":"pith_short_12","alias_value":"VJIBQPNWAF4Q","created_at":"2026-07-05T11:19:52.604824+00:00"},{"alias_kind":"pith_short_16","alias_value":"VJIBQPNWAF4QR34T","created_at":"2026-07-05T11:19:52.604824+00:00"},{"alias_kind":"pith_short_8","alias_value":"VJIBQPNW","created_at":"2026-07-05T11:19:52.604824+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06764","citing_title":"Cost-Effective Agent Harnesses for Abstract Reasoning and Generalization on ARC-AGI-1","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2604.18839","citing_title":"One Step Forward and K Steps Back: Better Reasoning with Denoising Recursion Models","ref_index":155,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK","json":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK.json","graph_json":"https://pith.science/api/pith-number/VJIBQPNWAF4QR34T6QK4RUDCRK/graph.json","events_json":"https://pith.science/api/pith-number/VJIBQPNWAF4QR34T6QK4RUDCRK/events.json","paper":"https://pith.science/paper/VJIBQPNW"},"agent_actions":{"view_html":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK","download_json":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK.json","view_paper":"https://pith.science/paper/VJIBQPNW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.07859&json=true","fetch_graph":"https://pith.science/api/pith-number/VJIBQPNWAF4QR34T6QK4RUDCRK/graph.json","fetch_events":"https://pith.science/api/pith-number/VJIBQPNWAF4QR34T6QK4RUDCRK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK/action/storage_attestation","attest_author":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK/action/author_attestation","sign_citation":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK/action/citation_signature","submit_replication":"https://pith.science/pith/VJIBQPNWAF4QR34T6QK4RUDCRK/action/replication_record"}},"created_at":"2026-07-05T11:19:52.604824+00:00","updated_at":"2026-07-05T11:19:52.604824+00:00"}