{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QCRIXGWFC7KS65JW2R52E7W7G2","short_pith_number":"pith:QCRIXGWF","schema_version":"1.0","canonical_sha256":"80a28b9ac517d52f7536d47ba27edf36b707f2f4c631d92294c11b1d4dc78208","source":{"kind":"arxiv","id":"2410.02703","version":2},"attestation_state":"computed","paper":{"title":"Selective Attention Improves Transformer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Matan Kalman, Yaniv Leviathan, Yossi Matias","submitted_at":"2024-10-03T17:27:30Z","abstract_excerpt":"Unneeded elements in the attention's context degrade performance. We introduce Selective Attention, a simple parameter-free change to the standard attention mechanism which reduces attention to unneeded elements. Selective attention consistently improves language modeling and downstream task performance in a variety of model sizes and context lengths. For example, transformers trained with the language modeling objective on C4 with selective attention perform language modeling equivalently to standard transformers with ~2X more heads and parameters in their attention modules. Selective attenti"},"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":"2410.02703","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-03T17:27:30Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"01cf89018f7aa01233eba097b382a67fad683c4d23aea828f8affb04760dea90","abstract_canon_sha256":"294a279c01343c2095ad5ccb58a47a99bbb7d15734e89a1c72f0bebd7ece87ee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:53:12.198494Z","signature_b64":"rjHpFFN+djBvOFWkcMhp1VF8YdpKoUSTkLV7JxSPggh56d4s6Kptonl3loLnaCZFMoDPLtg6T0SNpXDWF1b0CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"80a28b9ac517d52f7536d47ba27edf36b707f2f4c631d92294c11b1d4dc78208","last_reissued_at":"2026-07-05T10:53:12.198018Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:53:12.198018Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Selective Attention Improves Transformer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Matan Kalman, Yaniv Leviathan, Yossi Matias","submitted_at":"2024-10-03T17:27:30Z","abstract_excerpt":"Unneeded elements in the attention's context degrade performance. We introduce Selective Attention, a simple parameter-free change to the standard attention mechanism which reduces attention to unneeded elements. Selective attention consistently improves language modeling and downstream task performance in a variety of model sizes and context lengths. For example, transformers trained with the language modeling objective on C4 with selective attention perform language modeling equivalently to standard transformers with ~2X more heads and parameters in their attention modules. Selective attenti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02703","kind":"arxiv","version":2},"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/2410.02703/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":"2410.02703","created_at":"2026-07-05T10:53:12.198076+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.02703v2","created_at":"2026-07-05T10:53:12.198076+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02703","created_at":"2026-07-05T10:53:12.198076+00:00"},{"alias_kind":"pith_short_12","alias_value":"QCRIXGWFC7KS","created_at":"2026-07-05T10:53:12.198076+00:00"},{"alias_kind":"pith_short_16","alias_value":"QCRIXGWFC7KS65JW","created_at":"2026-07-05T10:53:12.198076+00:00"},{"alias_kind":"pith_short_8","alias_value":"QCRIXGWF","created_at":"2026-07-05T10:53:12.198076+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30429","citing_title":"Attention-based optimizer for symmetry finding","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20798","citing_title":"Most Transformer Modifications Still Do Not Transfer at 1-3B: A 2020-2026 Update to Narang et al. (2021) with Downstream Evaluation and a Noise Floor","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02568","citing_title":"StreamIndex: Memory-Bounded Compressed Sparse Attention via Streaming Top-k","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2","json":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2.json","graph_json":"https://pith.science/api/pith-number/QCRIXGWFC7KS65JW2R52E7W7G2/graph.json","events_json":"https://pith.science/api/pith-number/QCRIXGWFC7KS65JW2R52E7W7G2/events.json","paper":"https://pith.science/paper/QCRIXGWF"},"agent_actions":{"view_html":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2","download_json":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2.json","view_paper":"https://pith.science/paper/QCRIXGWF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.02703&json=true","fetch_graph":"https://pith.science/api/pith-number/QCRIXGWFC7KS65JW2R52E7W7G2/graph.json","fetch_events":"https://pith.science/api/pith-number/QCRIXGWFC7KS65JW2R52E7W7G2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2/action/storage_attestation","attest_author":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2/action/author_attestation","sign_citation":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2/action/citation_signature","submit_replication":"https://pith.science/pith/QCRIXGWFC7KS65JW2R52E7W7G2/action/replication_record"}},"created_at":"2026-07-05T10:53:12.198076+00:00","updated_at":"2026-07-05T10:53:12.198076+00:00"}