{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M5WSJ5VFCVHBSBMR6RHPRWNAVE","short_pith_number":"pith:M5WSJ5VF","schema_version":"1.0","canonical_sha256":"676d24f6a5154e190591f44ef8d9a0a9024e64d1063c937f9b8a5a528e73a2b7","source":{"kind":"arxiv","id":"2502.12033","version":1},"attestation_state":"computed","paper":{"title":"The geometry of BERT","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Giansalvo Cirrincione, Giorgia Ghione, Matteo Bonino","submitted_at":"2025-02-17T17:03:12Z","abstract_excerpt":"Transformer neural networks, particularly Bidirectional Encoder Representations from Transformers (BERT), have shown remarkable performance across various tasks such as classification, text summarization, and question answering. However, their internal mechanisms remain mathematically obscure, highlighting the need for greater explainability and interpretability. In this direction, this paper investigates the internal mechanisms of BERT proposing a novel perspective on the attention mechanism of BERT from a theoretical perspective. The analysis encompasses both local and global network behavio"},"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":"2502.12033","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T17:03:12Z","cross_cats_sorted":[],"title_canon_sha256":"f9f6af2b2ff75bcd7ffdcef3dbea6e7dff4cd07348dcb82ec4932adc3e3cd638","abstract_canon_sha256":"e35a256eea6d32dc3f4dd6a6d7cbccbbe3fda7e362c719c528a5cc374a352987"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:37.647687Z","signature_b64":"n5x96+vMLxSXEf78SZPhC0BN2gmcs+nGpDs99O2XQKbP760zia7hVpGGA7kITXJoLIgLW+bl9KayITzqSWeEAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"676d24f6a5154e190591f44ef8d9a0a9024e64d1063c937f9b8a5a528e73a2b7","last_reissued_at":"2026-07-05T10:15:37.647154Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:37.647154Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The geometry of BERT","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Giansalvo Cirrincione, Giorgia Ghione, Matteo Bonino","submitted_at":"2025-02-17T17:03:12Z","abstract_excerpt":"Transformer neural networks, particularly Bidirectional Encoder Representations from Transformers (BERT), have shown remarkable performance across various tasks such as classification, text summarization, and question answering. However, their internal mechanisms remain mathematically obscure, highlighting the need for greater explainability and interpretability. In this direction, this paper investigates the internal mechanisms of BERT proposing a novel perspective on the attention mechanism of BERT from a theoretical perspective. The analysis encompasses both local and global network behavio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12033","kind":"arxiv","version":1},"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/2502.12033/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":"2502.12033","created_at":"2026-07-05T10:15:37.647220+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.12033v1","created_at":"2026-07-05T10:15:37.647220+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12033","created_at":"2026-07-05T10:15:37.647220+00:00"},{"alias_kind":"pith_short_12","alias_value":"M5WSJ5VFCVHB","created_at":"2026-07-05T10:15:37.647220+00:00"},{"alias_kind":"pith_short_16","alias_value":"M5WSJ5VFCVHBSBMR","created_at":"2026-07-05T10:15:37.647220+00:00"},{"alias_kind":"pith_short_8","alias_value":"M5WSJ5VF","created_at":"2026-07-05T10:15:37.647220+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.23681","citing_title":"Rank, Head-Channel Non-Identifiability, and Symmetry Breaking: A Precise Analysis of Representational Collapse in Transformers","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10703","citing_title":"INCRT: An Incremental Transformer That Determines Its Own Architecture","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05217","citing_title":"On the Geometry of Positional Encodings in Transformers","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE","json":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE.json","graph_json":"https://pith.science/api/pith-number/M5WSJ5VFCVHBSBMR6RHPRWNAVE/graph.json","events_json":"https://pith.science/api/pith-number/M5WSJ5VFCVHBSBMR6RHPRWNAVE/events.json","paper":"https://pith.science/paper/M5WSJ5VF"},"agent_actions":{"view_html":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE","download_json":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE.json","view_paper":"https://pith.science/paper/M5WSJ5VF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.12033&json=true","fetch_graph":"https://pith.science/api/pith-number/M5WSJ5VFCVHBSBMR6RHPRWNAVE/graph.json","fetch_events":"https://pith.science/api/pith-number/M5WSJ5VFCVHBSBMR6RHPRWNAVE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE/action/storage_attestation","attest_author":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE/action/author_attestation","sign_citation":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE/action/citation_signature","submit_replication":"https://pith.science/pith/M5WSJ5VFCVHBSBMR6RHPRWNAVE/action/replication_record"}},"created_at":"2026-07-05T10:15:37.647220+00:00","updated_at":"2026-07-05T10:15:37.647220+00:00"}