{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:5WJF35HDBPSZVMNUXK2JZICRXS","short_pith_number":"pith:5WJF35HD","schema_version":"1.0","canonical_sha256":"ed925df4e30be59ab1b4bab49ca051bca0907bfe823fd23e2ae461cefb1a5512","source":{"kind":"arxiv","id":"2302.04542","version":1},"attestation_state":"computed","paper":{"title":"Efficient Attention via Control Variates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Chong Wang, Jianbo Yuan, Lingpeng Kong, Lin Zheng","submitted_at":"2023-02-09T10:16:20Z","abstract_excerpt":"Random-feature-based attention (RFA) is an efficient approximation of softmax attention with linear runtime and space complexity. However, the approximation gap between RFA and conventional softmax attention is not well studied. Built upon previous progress of RFA, we characterize this gap through the lens of control variates and show that RFA can be decomposed into a sum of multiple control variate estimators for each element in the sequence. This new framework reveals that exact softmax attention can be recovered from RFA by manipulating each control variate. Besides, it allows us to develop"},"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":"2302.04542","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-09T10:16:20Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"a85d45484a682b28dbc22a854ff78f95907bfacbe1e1815661810b1182f599d6","abstract_canon_sha256":"d096d93722eda513d2d643cb41573ec4f6fddd3179d690deb7fe1d93847dd9e8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:40:15.535247Z","signature_b64":"3RTzq/LnSufaFdbM3+SFKze7iM17/M2Z9tMDcOjj4TRJ4rbhGSONPZMAkeGDlUn9JqCFX5mjIlnSZuPoOwHhDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed925df4e30be59ab1b4bab49ca051bca0907bfe823fd23e2ae461cefb1a5512","last_reissued_at":"2026-07-05T05:40:15.534759Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:40:15.534759Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Attention via Control Variates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Chong Wang, Jianbo Yuan, Lingpeng Kong, Lin Zheng","submitted_at":"2023-02-09T10:16:20Z","abstract_excerpt":"Random-feature-based attention (RFA) is an efficient approximation of softmax attention with linear runtime and space complexity. However, the approximation gap between RFA and conventional softmax attention is not well studied. Built upon previous progress of RFA, we characterize this gap through the lens of control variates and show that RFA can be decomposed into a sum of multiple control variate estimators for each element in the sequence. This new framework reveals that exact softmax attention can be recovered from RFA by manipulating each control variate. Besides, it allows us to develop"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.04542","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/2302.04542/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":"2302.04542","created_at":"2026-07-05T05:40:15.534818+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.04542v1","created_at":"2026-07-05T05:40:15.534818+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.04542","created_at":"2026-07-05T05:40:15.534818+00:00"},{"alias_kind":"pith_short_12","alias_value":"5WJF35HDBPSZ","created_at":"2026-07-05T05:40:15.534818+00:00"},{"alias_kind":"pith_short_16","alias_value":"5WJF35HDBPSZVMNU","created_at":"2026-07-05T05:40:15.534818+00:00"},{"alias_kind":"pith_short_8","alias_value":"5WJF35HD","created_at":"2026-07-05T05:40:15.534818+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.09101","citing_title":"Feature Shift Localization Network","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS","json":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS.json","graph_json":"https://pith.science/api/pith-number/5WJF35HDBPSZVMNUXK2JZICRXS/graph.json","events_json":"https://pith.science/api/pith-number/5WJF35HDBPSZVMNUXK2JZICRXS/events.json","paper":"https://pith.science/paper/5WJF35HD"},"agent_actions":{"view_html":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS","download_json":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS.json","view_paper":"https://pith.science/paper/5WJF35HD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.04542&json=true","fetch_graph":"https://pith.science/api/pith-number/5WJF35HDBPSZVMNUXK2JZICRXS/graph.json","fetch_events":"https://pith.science/api/pith-number/5WJF35HDBPSZVMNUXK2JZICRXS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS/action/storage_attestation","attest_author":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS/action/author_attestation","sign_citation":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS/action/citation_signature","submit_replication":"https://pith.science/pith/5WJF35HDBPSZVMNUXK2JZICRXS/action/replication_record"}},"created_at":"2026-07-05T05:40:15.534818+00:00","updated_at":"2026-07-05T05:40:15.534818+00:00"}