{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ERLKSZQ5FICD5R25XAMARTHNRF","short_pith_number":"pith:ERLKSZQ5","schema_version":"1.0","canonical_sha256":"2456a9661d2a043ec75db81808cced894497f6e22f6b76f4b4fcf864ea792893","source":{"kind":"arxiv","id":"2409.13728","version":2},"attestation_state":"computed","paper":{"title":"Rule Extrapolation in Language Models: A Study of Compositional Generalization on OOD Prompts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Anna M\\'esz\\'aros, Ferenc Husz\\'ar, Patrik Reizinger, Szilvia Ujv\\'ary, Wieland Brendel","submitted_at":"2024-09-09T22:36:35Z","abstract_excerpt":"LLMs show remarkable emergent abilities, such as inferring concepts from presumably out-of-distribution prompts, known as in-context learning. Though this success is often attributed to the Transformer architecture, our systematic understanding is limited. In complex real-world data sets, even defining what is out-of-distribution is not obvious. To better understand the OOD behaviour of autoregressive LLMs, we focus on formal languages, which are defined by the intersection of rules. We define a new scenario of OOD compositional generalization, termed rule extrapolation. Rule extrapolation des"},"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":"2409.13728","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-09T22:36:35Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"16062416e8c70b5edb09c7563dfd59e6c0fca6403918b39b78365a87ccca7a96","abstract_canon_sha256":"e361b969565c5ba69c9fb078070202c178cb74ab1327a4b90e97b6c88079801b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:25:01.436797Z","signature_b64":"qOsaiIszCREq7+RzKhQu2WQQ6OixCUXf0jTeTHHondUGDD2t1Khad7u4a66CkSrSMAtBIRGZ7vYnODn83vMTCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2456a9661d2a043ec75db81808cced894497f6e22f6b76f4b4fcf864ea792893","last_reissued_at":"2026-07-05T09:25:01.436353Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:25:01.436353Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rule Extrapolation in Language Models: A Study of Compositional Generalization on OOD Prompts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Anna M\\'esz\\'aros, Ferenc Husz\\'ar, Patrik Reizinger, Szilvia Ujv\\'ary, Wieland Brendel","submitted_at":"2024-09-09T22:36:35Z","abstract_excerpt":"LLMs show remarkable emergent abilities, such as inferring concepts from presumably out-of-distribution prompts, known as in-context learning. Though this success is often attributed to the Transformer architecture, our systematic understanding is limited. In complex real-world data sets, even defining what is out-of-distribution is not obvious. To better understand the OOD behaviour of autoregressive LLMs, we focus on formal languages, which are defined by the intersection of rules. We define a new scenario of OOD compositional generalization, termed rule extrapolation. Rule extrapolation des"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.13728","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/2409.13728/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":"2409.13728","created_at":"2026-07-05T09:25:01.436414+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.13728v2","created_at":"2026-07-05T09:25:01.436414+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.13728","created_at":"2026-07-05T09:25:01.436414+00:00"},{"alias_kind":"pith_short_12","alias_value":"ERLKSZQ5FICD","created_at":"2026-07-05T09:25:01.436414+00:00"},{"alias_kind":"pith_short_16","alias_value":"ERLKSZQ5FICD5R25","created_at":"2026-07-05T09:25:01.436414+00:00"},{"alias_kind":"pith_short_8","alias_value":"ERLKSZQ5","created_at":"2026-07-05T09:25:01.436414+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF","json":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF.json","graph_json":"https://pith.science/api/pith-number/ERLKSZQ5FICD5R25XAMARTHNRF/graph.json","events_json":"https://pith.science/api/pith-number/ERLKSZQ5FICD5R25XAMARTHNRF/events.json","paper":"https://pith.science/paper/ERLKSZQ5"},"agent_actions":{"view_html":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF","download_json":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF.json","view_paper":"https://pith.science/paper/ERLKSZQ5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.13728&json=true","fetch_graph":"https://pith.science/api/pith-number/ERLKSZQ5FICD5R25XAMARTHNRF/graph.json","fetch_events":"https://pith.science/api/pith-number/ERLKSZQ5FICD5R25XAMARTHNRF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF/action/storage_attestation","attest_author":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF/action/author_attestation","sign_citation":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF/action/citation_signature","submit_replication":"https://pith.science/pith/ERLKSZQ5FICD5R25XAMARTHNRF/action/replication_record"}},"created_at":"2026-07-05T09:25:01.436414+00:00","updated_at":"2026-07-05T09:25:01.436414+00:00"}