{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7XNBJK4PHFKYZAKDA3M7FFEBRC","short_pith_number":"pith:7XNBJK4P","schema_version":"1.0","canonical_sha256":"fdda14ab8f39558c814306d9f29481889b88bd5b5542367db0a6e551a7630212","source":{"kind":"arxiv","id":"2607.07706","version":1},"attestation_state":"computed","paper":{"title":"The Key to Going Linear: Analysis-Driven Transformer Linearization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anna Kuzina, Babak Ehteshami Bejnordi, Paul N. Whatmough","submitted_at":"2026-07-08T17:59:09Z","abstract_excerpt":"The quadratic cost of causal self-attention severely bottlenecks long-context transformer inference. While numerous post hoc linearization pipelines exist, it is difficult to identify which components preserve model quality. This work isolates the effect of state update design in a strict frozen-backbone regime. We show that softmax relies on key-dependent, rank-1 orthogonal projections, elucidating why delta-style networks outperform purely gated accumulation. We identify a potential source of approximation errors and introduce structural interventions, specifically sink tokens, short convolu"},"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":"2607.07706","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-08T17:59:09Z","cross_cats_sorted":[],"title_canon_sha256":"3a460b98d87f39e363790d525d3a62549ec6265e8e39d50fe784896a6f4e0ff9","abstract_canon_sha256":"1f26cd29ff03fe2ef219f83d7087d2b7cac4e77c00d05b0ebfd1225edddf4bf7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:20:39.784377Z","signature_b64":"bfVZb1AvCNY1Cs3cE4xtPDna3cPQfAFCa05e57QiD9obmytZMmQL+9ozrvz0TiUdavoB40VyYyM+/VDhL9ncCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fdda14ab8f39558c814306d9f29481889b88bd5b5542367db0a6e551a7630212","last_reissued_at":"2026-07-09T01:20:39.783944Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:20:39.783944Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Key to Going Linear: Analysis-Driven Transformer Linearization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anna Kuzina, Babak Ehteshami Bejnordi, Paul N. Whatmough","submitted_at":"2026-07-08T17:59:09Z","abstract_excerpt":"The quadratic cost of causal self-attention severely bottlenecks long-context transformer inference. While numerous post hoc linearization pipelines exist, it is difficult to identify which components preserve model quality. This work isolates the effect of state update design in a strict frozen-backbone regime. We show that softmax relies on key-dependent, rank-1 orthogonal projections, elucidating why delta-style networks outperform purely gated accumulation. We identify a potential source of approximation errors and introduce structural interventions, specifically sink tokens, short convolu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.07706","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/2607.07706/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":"2607.07706","created_at":"2026-07-09T01:20:39.784005+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.07706v1","created_at":"2026-07-09T01:20:39.784005+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.07706","created_at":"2026-07-09T01:20:39.784005+00:00"},{"alias_kind":"pith_short_12","alias_value":"7XNBJK4PHFKY","created_at":"2026-07-09T01:20:39.784005+00:00"},{"alias_kind":"pith_short_16","alias_value":"7XNBJK4PHFKYZAKD","created_at":"2026-07-09T01:20:39.784005+00:00"},{"alias_kind":"pith_short_8","alias_value":"7XNBJK4P","created_at":"2026-07-09T01:20:39.784005+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/7XNBJK4PHFKYZAKDA3M7FFEBRC","json":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC.json","graph_json":"https://pith.science/api/pith-number/7XNBJK4PHFKYZAKDA3M7FFEBRC/graph.json","events_json":"https://pith.science/api/pith-number/7XNBJK4PHFKYZAKDA3M7FFEBRC/events.json","paper":"https://pith.science/paper/7XNBJK4P"},"agent_actions":{"view_html":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC","download_json":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC.json","view_paper":"https://pith.science/paper/7XNBJK4P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.07706&json=true","fetch_graph":"https://pith.science/api/pith-number/7XNBJK4PHFKYZAKDA3M7FFEBRC/graph.json","fetch_events":"https://pith.science/api/pith-number/7XNBJK4PHFKYZAKDA3M7FFEBRC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC/action/storage_attestation","attest_author":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC/action/author_attestation","sign_citation":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC/action/citation_signature","submit_replication":"https://pith.science/pith/7XNBJK4PHFKYZAKDA3M7FFEBRC/action/replication_record"}},"created_at":"2026-07-09T01:20:39.784005+00:00","updated_at":"2026-07-09T01:20:39.784005+00:00"}