{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Z5APCI36GCFTML5I6G3GEJZSH6","short_pith_number":"pith:Z5APCI36","schema_version":"1.0","canonical_sha256":"cf40f1237e308b362fa8f1b66227323f9b9b3f251c2b52625e28cf5d0008d3e0","source":{"kind":"arxiv","id":"2506.12284","version":2},"attestation_state":"computed","paper":{"title":"GrokAlign: Geometric Characterisation and Acceleration of Grokking","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ahmed Imtiaz Humayun, Randall Balestriero, Richard Baraniuk, Thomas Walker","submitted_at":"2025-06-14T00:16:11Z","abstract_excerpt":"A key challenge for the machine learning community is to understand and accelerate the training dynamics of deep networks that lead to delayed generalisation and emergent robustness to input perturbations, also known as grokking. Prior work has associated phenomena like delayed generalisation with the transition of a deep network from a linear to a feature learning regime, and emergent robustness with changes to the network's functional geometry, in particular the arrangement of the so-called linear regions in deep networks employing continuous piecewise affine nonlinearities. Here, we explain"},"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":"2506.12284","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-14T00:16:11Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"568c24baed188311ebf1fdcbb1c36f1c1f2eeae2bb786d784556af121939e7ed","abstract_canon_sha256":"4c639bd53f266edde8703f9479abaaf7aea49cea66826c7aa9e7308196efc21d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:45:58.855983Z","signature_b64":"jKxlAys1QlBGgGLH6adKZZVrow21EkA72gBIoPAI7xfiX3RuKV0W5VaBcLj+hTDx7uA37mzyqgjZPkbhHe4YBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cf40f1237e308b362fa8f1b66227323f9b9b3f251c2b52625e28cf5d0008d3e0","last_reissued_at":"2026-07-05T11:45:58.855432Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:45:58.855432Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GrokAlign: Geometric Characterisation and Acceleration of Grokking","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ahmed Imtiaz Humayun, Randall Balestriero, Richard Baraniuk, Thomas Walker","submitted_at":"2025-06-14T00:16:11Z","abstract_excerpt":"A key challenge for the machine learning community is to understand and accelerate the training dynamics of deep networks that lead to delayed generalisation and emergent robustness to input perturbations, also known as grokking. Prior work has associated phenomena like delayed generalisation with the transition of a deep network from a linear to a feature learning regime, and emergent robustness with changes to the network's functional geometry, in particular the arrangement of the so-called linear regions in deep networks employing continuous piecewise affine nonlinearities. Here, we explain"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.12284","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/2506.12284/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":"2506.12284","created_at":"2026-07-05T11:45:58.855492+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.12284v2","created_at":"2026-07-05T11:45:58.855492+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.12284","created_at":"2026-07-05T11:45:58.855492+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z5APCI36GCFT","created_at":"2026-07-05T11:45:58.855492+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z5APCI36GCFTML5I","created_at":"2026-07-05T11:45:58.855492+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z5APCI36","created_at":"2026-07-05T11:45:58.855492+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.04930","citing_title":"Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6","json":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6.json","graph_json":"https://pith.science/api/pith-number/Z5APCI36GCFTML5I6G3GEJZSH6/graph.json","events_json":"https://pith.science/api/pith-number/Z5APCI36GCFTML5I6G3GEJZSH6/events.json","paper":"https://pith.science/paper/Z5APCI36"},"agent_actions":{"view_html":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6","download_json":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6.json","view_paper":"https://pith.science/paper/Z5APCI36","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.12284&json=true","fetch_graph":"https://pith.science/api/pith-number/Z5APCI36GCFTML5I6G3GEJZSH6/graph.json","fetch_events":"https://pith.science/api/pith-number/Z5APCI36GCFTML5I6G3GEJZSH6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6/action/storage_attestation","attest_author":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6/action/author_attestation","sign_citation":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6/action/citation_signature","submit_replication":"https://pith.science/pith/Z5APCI36GCFTML5I6G3GEJZSH6/action/replication_record"}},"created_at":"2026-07-05T11:45:58.855492+00:00","updated_at":"2026-07-05T11:45:58.855492+00:00"}