{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6QJWXCMAYVREY2MEXPSWUVWCZE","short_pith_number":"pith:6QJWXCMA","schema_version":"1.0","canonical_sha256":"f4136b8980c5624c6984bbe56a56c2c93fef7fa98d0a6d897b0ffd8383078d64","source":{"kind":"arxiv","id":"2605.17084","version":1},"attestation_state":"computed","paper":{"title":"Scale Determines Whether Language Models Organize Representation Geometry for Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Weilun Xu","submitted_at":"2026-05-16T17:01:17Z","abstract_excerpt":"In language models, what a representation encodes is determined by the geometry of its representation space: distances, not activations, carry meaning. Existing tools characterize the shape of this geometry but do not ask what that shape is organized for. We introduce Subspace PGA, a metric that tests whether a layer's distance structure aligns with the readout subspace of the unembedding matrix $W_U$ more than with random subspaces of equal size. Across seven Pythia models (70M--6.9B) and three cross-family models, intermediate geometry is significantly organized for prediction (peak $z = 9$-"},"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":"2605.17084","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-16T17:01:17Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"5e5ac31f5da0330896a2c6941edec93177dfa0778471e13d734fbc7981d228f1","abstract_canon_sha256":"f4464a7d0409ae1fb93d6f9088c17fd7452cbb6e3ef0ae29ad963871d3bee47b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:39.569249Z","signature_b64":"Y19xk16PyaRGwjhsCyIwiiT4jzUYxUGlAvGcuUDwWNX3BwFBQ2PIIwzPC6QKW7niMN7YTf+Db2i3stkA2AOSDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f4136b8980c5624c6984bbe56a56c2c93fef7fa98d0a6d897b0ffd8383078d64","last_reissued_at":"2026-05-20T00:03:39.568786Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:39.568786Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scale Determines Whether Language Models Organize Representation Geometry for Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Weilun Xu","submitted_at":"2026-05-16T17:01:17Z","abstract_excerpt":"In language models, what a representation encodes is determined by the geometry of its representation space: distances, not activations, carry meaning. Existing tools characterize the shape of this geometry but do not ask what that shape is organized for. We introduce Subspace PGA, a metric that tests whether a layer's distance structure aligns with the readout subspace of the unembedding matrix $W_U$ more than with random subspaces of equal size. Across seven Pythia models (70M--6.9B) and three cross-family models, intermediate geometry is significantly organized for prediction (peak $z = 9$-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.17084","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/2605.17084/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:33:23.805991Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T22:21:57.742143Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"44e09bc5468d8d8b56062ff1db82fecefc59c7b4608e526ff597fdadd1982ab3"},"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":"2605.17084","created_at":"2026-05-20T00:03:39.568849+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17084v1","created_at":"2026-05-20T00:03:39.568849+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17084","created_at":"2026-05-20T00:03:39.568849+00:00"},{"alias_kind":"pith_short_12","alias_value":"6QJWXCMAYVRE","created_at":"2026-05-20T00:03:39.568849+00:00"},{"alias_kind":"pith_short_16","alias_value":"6QJWXCMAYVREY2ME","created_at":"2026-05-20T00:03:39.568849+00:00"},{"alias_kind":"pith_short_8","alias_value":"6QJWXCMA","created_at":"2026-05-20T00:03:39.568849+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE","json":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE.json","graph_json":"https://pith.science/api/pith-number/6QJWXCMAYVREY2MEXPSWUVWCZE/graph.json","events_json":"https://pith.science/api/pith-number/6QJWXCMAYVREY2MEXPSWUVWCZE/events.json","paper":"https://pith.science/paper/6QJWXCMA"},"agent_actions":{"view_html":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE","download_json":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE.json","view_paper":"https://pith.science/paper/6QJWXCMA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17084&json=true","fetch_graph":"https://pith.science/api/pith-number/6QJWXCMAYVREY2MEXPSWUVWCZE/graph.json","fetch_events":"https://pith.science/api/pith-number/6QJWXCMAYVREY2MEXPSWUVWCZE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE/action/storage_attestation","attest_author":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE/action/author_attestation","sign_citation":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE/action/citation_signature","submit_replication":"https://pith.science/pith/6QJWXCMAYVREY2MEXPSWUVWCZE/action/replication_record"}},"created_at":"2026-05-20T00:03:39.568849+00:00","updated_at":"2026-05-20T00:03:39.568849+00:00"}