{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:CDNVENGAHWF6CD5ZWFICY3EJRT","short_pith_number":"pith:CDNVENGA","schema_version":"1.0","canonical_sha256":"10db5234c03d8be10fb9b1502c6c898cfb421eaea5f0cadab1c67b1177460a95","source":{"kind":"arxiv","id":"2602.15029","version":3},"attestation_state":"computed","paper":{"title":"Symmetry in language statistics shapes the geometry of model representations","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.CL"],"primary_cat":"cs.LG","authors_text":"Andres Nava, Daniel J. Korchinski, Dhruva Karkada, Matthieu Wyart, Yasaman Bahri","submitted_at":"2026-02-16T18:59:55Z","abstract_excerpt":"The internal representations learned by language models consistently exhibit striking geometric structure: calendar months organize into a circle, historical years form a smooth one-dimensional manifold, and cities' latitudes and longitudes can be decoded using a linear probe. To explain this neural code, we first show that language statistics exhibit translation symmetry (for example, the frequency with which any two months co-occur in text depends only on the time interval between them). We prove that this symmetry governs these geometric structures in high-dimensional word embedding models,"},"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":"2602.15029","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-16T18:59:55Z","cross_cats_sorted":["cond-mat.dis-nn","cs.CL"],"title_canon_sha256":"cdb477d6e3431592fa4c69080e0fb6d016b1454f0edb4d8473f383a99aaa42b8","abstract_canon_sha256":"ca499596cd83a75d852b6a30dc0dda7c96e52911c565fa4a53ff2c41f9a1c433"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T00:17:10.354207Z","signature_b64":"p+7qXTpgqFtwPkeEMD/WkOGzw5YlrStaJVGOot7T/B1Z/5ESt5XNRnTKk3oXAsUA9ZeylDzpe78Eu15yAUVuCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"10db5234c03d8be10fb9b1502c6c898cfb421eaea5f0cadab1c67b1177460a95","last_reissued_at":"2026-07-01T00:17:10.353703Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T00:17:10.353703Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Symmetry in language statistics shapes the geometry of model representations","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.CL"],"primary_cat":"cs.LG","authors_text":"Andres Nava, Daniel J. Korchinski, Dhruva Karkada, Matthieu Wyart, Yasaman Bahri","submitted_at":"2026-02-16T18:59:55Z","abstract_excerpt":"The internal representations learned by language models consistently exhibit striking geometric structure: calendar months organize into a circle, historical years form a smooth one-dimensional manifold, and cities' latitudes and longitudes can be decoded using a linear probe. To explain this neural code, we first show that language statistics exhibit translation symmetry (for example, the frequency with which any two months co-occur in text depends only on the time interval between them). We prove that this symmetry governs these geometric structures in high-dimensional word embedding models,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.15029","kind":"arxiv","version":3},"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/2602.15029/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":"2602.15029","created_at":"2026-07-01T00:17:10.353761+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.15029v3","created_at":"2026-07-01T00:17:10.353761+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.15029","created_at":"2026-07-01T00:17:10.353761+00:00"},{"alias_kind":"pith_short_12","alias_value":"CDNVENGAHWF6","created_at":"2026-07-01T00:17:10.353761+00:00"},{"alias_kind":"pith_short_16","alias_value":"CDNVENGAHWF6CD5Z","created_at":"2026-07-01T00:17:10.353761+00:00"},{"alias_kind":"pith_short_8","alias_value":"CDNVENGA","created_at":"2026-07-01T00:17:10.353761+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":10,"sample":[{"citing_arxiv_id":"2605.17482","citing_title":"RSD: Moving Local Triangular Charts for Auditing Language-Model Hidden States","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.27970","citing_title":"Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2606.27981","citing_title":"ToxiREX: A Dataset on Toxic REasoning in ConteXt","ref_index":81,"is_internal_anchor":true},{"citing_arxiv_id":"2605.23821","citing_title":"Hierarchical Concept Geometry in Language Models Emerges from Word Co-occurrence","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21486","citing_title":"Quantifying Hyperparameter Transfer and the Importance of Embedding Layer Learning Rate","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17482","citing_title":"RSD: Moving Local Triangular Charts for Auditing Language-Model Hidden States","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18537","citing_title":"Probing for Representation Manifolds in Superposition","ref_index":104,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12756","citing_title":"Uncovering Symmetry Transfer in Large Language Models via Layer-Peeled Optimization","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2604.21691","citing_title":"There Will Be a Scientific Theory of Deep Learning","ref_index":239,"is_internal_anchor":true},{"citing_arxiv_id":"2604.20817","citing_title":"Convergent Evolution: How Different Language Models Learn Similar Number Representations","ref_index":37,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT","json":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT.json","graph_json":"https://pith.science/api/pith-number/CDNVENGAHWF6CD5ZWFICY3EJRT/graph.json","events_json":"https://pith.science/api/pith-number/CDNVENGAHWF6CD5ZWFICY3EJRT/events.json","paper":"https://pith.science/paper/CDNVENGA"},"agent_actions":{"view_html":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT","download_json":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT.json","view_paper":"https://pith.science/paper/CDNVENGA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.15029&json=true","fetch_graph":"https://pith.science/api/pith-number/CDNVENGAHWF6CD5ZWFICY3EJRT/graph.json","fetch_events":"https://pith.science/api/pith-number/CDNVENGAHWF6CD5ZWFICY3EJRT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT/action/storage_attestation","attest_author":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT/action/author_attestation","sign_citation":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT/action/citation_signature","submit_replication":"https://pith.science/pith/CDNVENGAHWF6CD5ZWFICY3EJRT/action/replication_record"}},"created_at":"2026-07-01T00:17:10.353761+00:00","updated_at":"2026-07-01T00:17:10.353761+00:00"}