{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:72XFG6GRBL64SH6C4UINOZ6DRD","short_pith_number":"pith:72XFG6GR","schema_version":"1.0","canonical_sha256":"feae5378d10afdc91fc2e510d767c388c018d0893f9a665ad0bfb29369f1650c","source":{"kind":"arxiv","id":"2204.02892","version":4},"attestation_state":"computed","paper":{"title":"Sub-Task Decomposition Enables Learning in Sequence to Sequence Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Amnon Shashua, Noam Wies, Yoav Levine","submitted_at":"2022-04-06T15:16:27Z","abstract_excerpt":"The field of Natural Language Processing has experienced a dramatic leap in capabilities with the recent introduction of huge Language Models. Despite this success, natural language problems that involve several compounded steps are still practically unlearnable, even by the largest LMs. This complies with experimental failures for end-to-end learning of composite problems that were demonstrated in a variety of domains. An effective mitigation is to introduce intermediate supervision for solving sub-tasks of the compounded problem. Recently, several works have demonstrated high gains by taking"},"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":"2204.02892","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-04-06T15:16:27Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d70b9dcafdbd8954a16289fbde43a4b2b22894d77d895d2e9dec4d8be5192475","abstract_canon_sha256":"280651ceb8cc2aedd9a353f16619acfc6efead484daf303e15d391cfcf0f13c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:42:00.773181Z","signature_b64":"mqOneUdbsAR5CVSNr658lX7TexhaMMhgJkJGY4ZOGwytSQCHjziMaYkVWngFa1Z8rig/jiutE6HNKoCigZKKAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"feae5378d10afdc91fc2e510d767c388c018d0893f9a665ad0bfb29369f1650c","last_reissued_at":"2026-07-05T05:42:00.772744Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:42:00.772744Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sub-Task Decomposition Enables Learning in Sequence to Sequence Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Amnon Shashua, Noam Wies, Yoav Levine","submitted_at":"2022-04-06T15:16:27Z","abstract_excerpt":"The field of Natural Language Processing has experienced a dramatic leap in capabilities with the recent introduction of huge Language Models. Despite this success, natural language problems that involve several compounded steps are still practically unlearnable, even by the largest LMs. This complies with experimental failures for end-to-end learning of composite problems that were demonstrated in a variety of domains. An effective mitigation is to introduce intermediate supervision for solving sub-tasks of the compounded problem. Recently, several works have demonstrated high gains by taking"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2204.02892","kind":"arxiv","version":4},"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/2204.02892/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":"2204.02892","created_at":"2026-07-05T05:42:00.772800+00:00"},{"alias_kind":"arxiv_version","alias_value":"2204.02892v4","created_at":"2026-07-05T05:42:00.772800+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2204.02892","created_at":"2026-07-05T05:42:00.772800+00:00"},{"alias_kind":"pith_short_12","alias_value":"72XFG6GRBL64","created_at":"2026-07-05T05:42:00.772800+00:00"},{"alias_kind":"pith_short_16","alias_value":"72XFG6GRBL64SH6C","created_at":"2026-07-05T05:42:00.772800+00:00"},{"alias_kind":"pith_short_8","alias_value":"72XFG6GR","created_at":"2026-07-05T05:42:00.772800+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06720","citing_title":"When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven Reasoning","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.28600","citing_title":"Transformers Provably Learn to Internalize Chain-of-Thought","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00183","citing_title":"Agentic Transformers Provably Learn to Search via Reinforcement Learning","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18327","citing_title":"PARM: Pipeline-Adapted Reward Model","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD","json":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD.json","graph_json":"https://pith.science/api/pith-number/72XFG6GRBL64SH6C4UINOZ6DRD/graph.json","events_json":"https://pith.science/api/pith-number/72XFG6GRBL64SH6C4UINOZ6DRD/events.json","paper":"https://pith.science/paper/72XFG6GR"},"agent_actions":{"view_html":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD","download_json":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD.json","view_paper":"https://pith.science/paper/72XFG6GR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2204.02892&json=true","fetch_graph":"https://pith.science/api/pith-number/72XFG6GRBL64SH6C4UINOZ6DRD/graph.json","fetch_events":"https://pith.science/api/pith-number/72XFG6GRBL64SH6C4UINOZ6DRD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD/action/storage_attestation","attest_author":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD/action/author_attestation","sign_citation":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD/action/citation_signature","submit_replication":"https://pith.science/pith/72XFG6GRBL64SH6C4UINOZ6DRD/action/replication_record"}},"created_at":"2026-07-05T05:42:00.772800+00:00","updated_at":"2026-07-05T05:42:00.772800+00:00"}