{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:562J2F4A3B7OQZPUAR5YVB3V6X","short_pith_number":"pith:562J2F4A","schema_version":"1.0","canonical_sha256":"efb49d1780d87ee865f4047b8a8775f5e2922a1479ac2678f79b65795b3f12d9","source":{"kind":"arxiv","id":"2312.06581","version":2},"attestation_state":"computed","paper":{"title":"Grokking Group Multiplication with Cosets","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","math.RT"],"primary_cat":"cs.LG","authors_text":"Dashiell Stander, Honglu Fan, Qinan Yu, Stella Biderman","submitted_at":"2023-12-11T18:12:18Z","abstract_excerpt":"The complex and unpredictable nature of deep neural networks prevents their safe use in many high-stakes applications. There have been many techniques developed to interpret deep neural networks, but all have substantial limitations. Algorithmic tasks have proven to be a fruitful test ground for interpreting a neural network end-to-end. Building on previous work, we completely reverse engineer fully connected one-hidden layer networks that have ``grokked'' the arithmetic of the permutation groups $S_5$ and $S_6$. The models discover the true subgroup structure of the full group and converge on"},"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":"2312.06581","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-12-11T18:12:18Z","cross_cats_sorted":["cs.AI","math.RT"],"title_canon_sha256":"e8ff83b563dadacc7cd73bca935bebc93b6588bb6a45c939eb6e213607f46da7","abstract_canon_sha256":"a39ae01349e8c7c88b6f75bf01dbff13b394b1af0029a579f1ebcda7d0a81a6e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:32:33.739640Z","signature_b64":"9NBbuBzl04Qay6RZnjWysiDagVwSR4cLWSxt1DAcCGnHnFD+U0zT3ibqOkOQcwRfGZp7MRTilFnD5x1cADFTBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"efb49d1780d87ee865f4047b8a8775f5e2922a1479ac2678f79b65795b3f12d9","last_reissued_at":"2026-07-05T08:32:33.739115Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:32:33.739115Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Grokking Group Multiplication with Cosets","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","math.RT"],"primary_cat":"cs.LG","authors_text":"Dashiell Stander, Honglu Fan, Qinan Yu, Stella Biderman","submitted_at":"2023-12-11T18:12:18Z","abstract_excerpt":"The complex and unpredictable nature of deep neural networks prevents their safe use in many high-stakes applications. There have been many techniques developed to interpret deep neural networks, but all have substantial limitations. Algorithmic tasks have proven to be a fruitful test ground for interpreting a neural network end-to-end. Building on previous work, we completely reverse engineer fully connected one-hidden layer networks that have ``grokked'' the arithmetic of the permutation groups $S_5$ and $S_6$. The models discover the true subgroup structure of the full group and converge on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.06581","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/2312.06581/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":"2312.06581","created_at":"2026-07-05T08:32:33.739175+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.06581v2","created_at":"2026-07-05T08:32:33.739175+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.06581","created_at":"2026-07-05T08:32:33.739175+00:00"},{"alias_kind":"pith_short_12","alias_value":"562J2F4A3B7O","created_at":"2026-07-05T08:32:33.739175+00:00"},{"alias_kind":"pith_short_16","alias_value":"562J2F4A3B7OQZPU","created_at":"2026-07-05T08:32:33.739175+00:00"},{"alias_kind":"pith_short_8","alias_value":"562J2F4A","created_at":"2026-07-05T08:32:33.739175+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.12717","citing_title":"Learning the symmetric group: large from small","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06256","citing_title":"Spectral Edge Dynamics Reveal Functional Modes of Learning","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X","json":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X.json","graph_json":"https://pith.science/api/pith-number/562J2F4A3B7OQZPUAR5YVB3V6X/graph.json","events_json":"https://pith.science/api/pith-number/562J2F4A3B7OQZPUAR5YVB3V6X/events.json","paper":"https://pith.science/paper/562J2F4A"},"agent_actions":{"view_html":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X","download_json":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X.json","view_paper":"https://pith.science/paper/562J2F4A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.06581&json=true","fetch_graph":"https://pith.science/api/pith-number/562J2F4A3B7OQZPUAR5YVB3V6X/graph.json","fetch_events":"https://pith.science/api/pith-number/562J2F4A3B7OQZPUAR5YVB3V6X/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X/action/timestamp_anchor","attest_storage":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X/action/storage_attestation","attest_author":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X/action/author_attestation","sign_citation":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X/action/citation_signature","submit_replication":"https://pith.science/pith/562J2F4A3B7OQZPUAR5YVB3V6X/action/replication_record"}},"created_at":"2026-07-05T08:32:33.739175+00:00","updated_at":"2026-07-05T08:32:33.739175+00:00"}