{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PQRY6IH42X4DUBDGR55SWKB3EF","short_pith_number":"pith:PQRY6IH4","schema_version":"1.0","canonical_sha256":"7c238f20fcd5f83a04668f7b2b283b217ee9768f5d3cf584e6da9b0c4b03185e","source":{"kind":"arxiv","id":"2409.19808","version":2},"attestation_state":"computed","paper":{"title":"Can Models Learn Skill Composition from Examples?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anirudh Goyal, Dingli Yu, Haoyu Zhao, Sanjeev Arora, Simran Kaur","submitted_at":"2024-09-29T22:14:02Z","abstract_excerpt":"As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with com"},"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":"2409.19808","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-29T22:14:02Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"e94080b9470e31affdf1e3fff8c52f00f3ff9ae0eef0c7c03ed70af70b385017","abstract_canon_sha256":"72a82932a8dace748577db242df5d90d1d27f633b94774215a30b8468bea75fd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:02:52.660090Z","signature_b64":"CQ1uDLEr22HQEDvxay1m8xfZWNWhm8qA2mvb8H8QWF0a8mQEUnMFFyI9Qazgl/2vXicyLHp+Vjppvh7nMNDMDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c238f20fcd5f83a04668f7b2b283b217ee9768f5d3cf584e6da9b0c4b03185e","last_reissued_at":"2026-07-05T10:02:52.659530Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:02:52.659530Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can Models Learn Skill Composition from Examples?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anirudh Goyal, Dingli Yu, Haoyu Zhao, Sanjeev Arora, Simran Kaur","submitted_at":"2024-09-29T22:14:02Z","abstract_excerpt":"As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with com"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.19808","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/2409.19808/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":"2409.19808","created_at":"2026-07-05T10:02:52.659595+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.19808v2","created_at":"2026-07-05T10:02:52.659595+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.19808","created_at":"2026-07-05T10:02:52.659595+00:00"},{"alias_kind":"pith_short_12","alias_value":"PQRY6IH42X4D","created_at":"2026-07-05T10:02:52.659595+00:00"},{"alias_kind":"pith_short_16","alias_value":"PQRY6IH42X4DUBDG","created_at":"2026-07-05T10:02:52.659595+00:00"},{"alias_kind":"pith_short_8","alias_value":"PQRY6IH4","created_at":"2026-07-05T10:02:52.659595+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.05139","citing_title":"Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning","ref_index":67,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF","json":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF.json","graph_json":"https://pith.science/api/pith-number/PQRY6IH42X4DUBDGR55SWKB3EF/graph.json","events_json":"https://pith.science/api/pith-number/PQRY6IH42X4DUBDGR55SWKB3EF/events.json","paper":"https://pith.science/paper/PQRY6IH4"},"agent_actions":{"view_html":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF","download_json":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF.json","view_paper":"https://pith.science/paper/PQRY6IH4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.19808&json=true","fetch_graph":"https://pith.science/api/pith-number/PQRY6IH42X4DUBDGR55SWKB3EF/graph.json","fetch_events":"https://pith.science/api/pith-number/PQRY6IH42X4DUBDGR55SWKB3EF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF/action/storage_attestation","attest_author":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF/action/author_attestation","sign_citation":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF/action/citation_signature","submit_replication":"https://pith.science/pith/PQRY6IH42X4DUBDGR55SWKB3EF/action/replication_record"}},"created_at":"2026-07-05T10:02:52.659595+00:00","updated_at":"2026-07-05T10:02:52.659595+00:00"}