{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:D5RZTBE3RDAQGKLKKVGTHNFJCC","short_pith_number":"pith:D5RZTBE3","schema_version":"1.0","canonical_sha256":"1f6399849b88c103296a554d33b4a9109892ad33f5aa35d3ac2a34a715e334be","source":{"kind":"arxiv","id":"2110.04169","version":2},"attestation_state":"computed","paper":{"title":"Iterative Decoding for Compositional Generalization in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Joshua Ainslie, Luana Ruiz, Santiago Onta\\~n\\'on","submitted_at":"2021-10-08T14:52:25Z","abstract_excerpt":"Deep learning models generalize well to in-distribution data but struggle to generalize compositionally, i.e., to combine a set of learned primitives to solve more complex tasks. In sequence-to-sequence (seq2seq) learning, transformers are often unable to predict correct outputs for longer examples than those seen at training. This paper introduces iterative decoding, an alternative to seq2seq that (i) improves transformer compositional generalization in the PCFG and Cartesian product datasets and (ii) evidences that, in these datasets, seq2seq transformers do not learn iterations that are not"},"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":"2110.04169","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-10-08T14:52:25Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0acb1be8bf42ad4eb94b489fb47b8fb907aab7404034f4e9ff92f89bc3a58c3f","abstract_canon_sha256":"00da6d4de8a0fd1a9fd824982a6e27c3c1e92fa34917d653de6d578f38259f7f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:39:35.581028Z","signature_b64":"0E2TFh86qUSyYhbm5sIyV8hxweZVs0QaDaEm4CkDEV+C8kcqJOQtGrDvHfmTcOUx9jgcqe1L6BV95dHOGNoOAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f6399849b88c103296a554d33b4a9109892ad33f5aa35d3ac2a34a715e334be","last_reissued_at":"2026-07-05T03:39:35.580503Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:39:35.580503Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Iterative Decoding for Compositional Generalization in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Joshua Ainslie, Luana Ruiz, Santiago Onta\\~n\\'on","submitted_at":"2021-10-08T14:52:25Z","abstract_excerpt":"Deep learning models generalize well to in-distribution data but struggle to generalize compositionally, i.e., to combine a set of learned primitives to solve more complex tasks. In sequence-to-sequence (seq2seq) learning, transformers are often unable to predict correct outputs for longer examples than those seen at training. This paper introduces iterative decoding, an alternative to seq2seq that (i) improves transformer compositional generalization in the PCFG and Cartesian product datasets and (ii) evidences that, in these datasets, seq2seq transformers do not learn iterations that are not"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.04169","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/2110.04169/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":"2110.04169","created_at":"2026-07-05T03:39:35.580570+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.04169v2","created_at":"2026-07-05T03:39:35.580570+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.04169","created_at":"2026-07-05T03:39:35.580570+00:00"},{"alias_kind":"pith_short_12","alias_value":"D5RZTBE3RDAQ","created_at":"2026-07-05T03:39:35.580570+00:00"},{"alias_kind":"pith_short_16","alias_value":"D5RZTBE3RDAQGKLK","created_at":"2026-07-05T03:39:35.580570+00:00"},{"alias_kind":"pith_short_8","alias_value":"D5RZTBE3","created_at":"2026-07-05T03:39:35.580570+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/D5RZTBE3RDAQGKLKKVGTHNFJCC","json":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC.json","graph_json":"https://pith.science/api/pith-number/D5RZTBE3RDAQGKLKKVGTHNFJCC/graph.json","events_json":"https://pith.science/api/pith-number/D5RZTBE3RDAQGKLKKVGTHNFJCC/events.json","paper":"https://pith.science/paper/D5RZTBE3"},"agent_actions":{"view_html":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC","download_json":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC.json","view_paper":"https://pith.science/paper/D5RZTBE3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.04169&json=true","fetch_graph":"https://pith.science/api/pith-number/D5RZTBE3RDAQGKLKKVGTHNFJCC/graph.json","fetch_events":"https://pith.science/api/pith-number/D5RZTBE3RDAQGKLKKVGTHNFJCC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC/action/storage_attestation","attest_author":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC/action/author_attestation","sign_citation":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC/action/citation_signature","submit_replication":"https://pith.science/pith/D5RZTBE3RDAQGKLKKVGTHNFJCC/action/replication_record"}},"created_at":"2026-07-05T03:39:35.580570+00:00","updated_at":"2026-07-05T03:39:35.580570+00:00"}