{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3MSQGH6GBJCBMAK2CCAC2UQCUF","short_pith_number":"pith:3MSQGH6G","schema_version":"1.0","canonical_sha256":"db25031fc60a4416015a10802d5202a15171e59e61e98b67ab2048bf5c9ce3e2","source":{"kind":"arxiv","id":"2602.03655","version":2},"attestation_state":"computed","paper":{"title":"Sequential Group Composition: A Window into the Mechanics of Deep Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adele Myers, Daniel Kunin, Francisco Acosta, Giovanni Luca Marchetti, Nina Miolane","submitted_at":"2026-02-03T15:36:25Z","abstract_excerpt":"How do neural networks trained over sequences acquire the ability to perform structured operations, such as arithmetic, geometric, and algorithmic computation? To gain insight into this question, we introduce the sequential group composition task. In this task, networks receive a sequence of elements from a finite group encoded in a real vector space and must predict their cumulative product. This task can be order-sensitive and cannot be solved by a linear model. Our analysis isolates the roles of the group structure, encoding statistics, and sequence length in shaping learning. We prove that"},"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":"2602.03655","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-03T15:36:25Z","cross_cats_sorted":[],"title_canon_sha256":"102dc46103922d191d4d22c700047a466917937416d9edb2fc8d279e3526e08f","abstract_canon_sha256":"5bfafcae1bff3c2ae22f98494af85af7e26f2188db3a5a94534b720614a20e4d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:02:32.015753Z","signature_b64":"m85drEpJb7JBjHp9eUIJkK08N0YmhKXLhgikkNAMpf0+OS0MLhT8J2Y5sU5YP9UHJr0R96URbwXgWNA1Br1dAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"db25031fc60a4416015a10802d5202a15171e59e61e98b67ab2048bf5c9ce3e2","last_reissued_at":"2026-06-01T01:02:32.014686Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:02:32.014686Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sequential Group Composition: A Window into the Mechanics of Deep Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adele Myers, Daniel Kunin, Francisco Acosta, Giovanni Luca Marchetti, Nina Miolane","submitted_at":"2026-02-03T15:36:25Z","abstract_excerpt":"How do neural networks trained over sequences acquire the ability to perform structured operations, such as arithmetic, geometric, and algorithmic computation? To gain insight into this question, we introduce the sequential group composition task. In this task, networks receive a sequence of elements from a finite group encoded in a real vector space and must predict their cumulative product. This task can be order-sensitive and cannot be solved by a linear model. Our analysis isolates the roles of the group structure, encoding statistics, and sequence length in shaping learning. We prove that"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.03655","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/2602.03655/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":"2602.03655","created_at":"2026-06-01T01:02:32.014830+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.03655v2","created_at":"2026-06-01T01:02:32.014830+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.03655","created_at":"2026-06-01T01:02:32.014830+00:00"},{"alias_kind":"pith_short_12","alias_value":"3MSQGH6GBJCB","created_at":"2026-06-01T01:02:32.014830+00:00"},{"alias_kind":"pith_short_16","alias_value":"3MSQGH6GBJCBMAK2","created_at":"2026-06-01T01:02:32.014830+00:00"},{"alias_kind":"pith_short_8","alias_value":"3MSQGH6G","created_at":"2026-06-01T01:02:32.014830+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.05115","citing_title":"Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01148","citing_title":"Arithmetic in the Wild: Llama uses Base-10 Addition to Reason About Cyclic Concepts","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2604.21691","citing_title":"There Will Be a Scientific Theory of Deep Learning","ref_index":218,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF","json":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF.json","graph_json":"https://pith.science/api/pith-number/3MSQGH6GBJCBMAK2CCAC2UQCUF/graph.json","events_json":"https://pith.science/api/pith-number/3MSQGH6GBJCBMAK2CCAC2UQCUF/events.json","paper":"https://pith.science/paper/3MSQGH6G"},"agent_actions":{"view_html":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF","download_json":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF.json","view_paper":"https://pith.science/paper/3MSQGH6G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.03655&json=true","fetch_graph":"https://pith.science/api/pith-number/3MSQGH6GBJCBMAK2CCAC2UQCUF/graph.json","fetch_events":"https://pith.science/api/pith-number/3MSQGH6GBJCBMAK2CCAC2UQCUF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF/action/storage_attestation","attest_author":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF/action/author_attestation","sign_citation":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF/action/citation_signature","submit_replication":"https://pith.science/pith/3MSQGH6GBJCBMAK2CCAC2UQCUF/action/replication_record"}},"created_at":"2026-06-01T01:02:32.014830+00:00","updated_at":"2026-06-01T01:02:32.014830+00:00"}