{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:5XWW4GZF435KLLFAMR5A26NVT2","short_pith_number":"pith:5XWW4GZF","schema_version":"1.0","canonical_sha256":"eded6e1b25e6faa5aca0647a0d79b59e9746c5b0d219e71b782627a45ffb03bd","source":{"kind":"arxiv","id":"2305.14380","version":2},"attestation_state":"computed","paper":{"title":"Finding the Pillars of Strength for Multi-Head Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Erik Cambria, Han Lei, Jinjie Ni, Rui Mao, Zonglin Yang","submitted_at":"2023-05-22T03:44:44Z","abstract_excerpt":"Recent studies have revealed some issues of Multi-Head Attention (MHA), e.g., redundancy and over-parameterization. Specifically, the heads of MHA were originally designed to attend to information from different representation subspaces, whereas prior studies found that some attention heads likely learn similar features and can be pruned without harming performance. Inspired by the minimum-redundancy feature selection, we assume that focusing on the most representative and distinctive features with minimum resources can mitigate the above issues and lead to more effective and efficient MHAs. I"},"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":"2305.14380","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-05-22T03:44:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"d4e1c24d717018adfc800874e6e3de05ca5daf66392fe65151bad85092e9b8d5","abstract_canon_sha256":"083fd53bdfc0e2cb4c46f60612e82ebca8a1fb170a9a7459f932e95001b48467"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:01:01.754522Z","signature_b64":"l6PHwH/Os2lKfeFTm/FL1WhAjZwpUVi3mN83kafp0PEqFuKzzV8TGtrYYFDj6sstpDgvitYjoBj7fDg7KRg+Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eded6e1b25e6faa5aca0647a0d79b59e9746c5b0d219e71b782627a45ffb03bd","last_reissued_at":"2026-07-05T07:01:01.754017Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:01:01.754017Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Finding the Pillars of Strength for Multi-Head Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Erik Cambria, Han Lei, Jinjie Ni, Rui Mao, Zonglin Yang","submitted_at":"2023-05-22T03:44:44Z","abstract_excerpt":"Recent studies have revealed some issues of Multi-Head Attention (MHA), e.g., redundancy and over-parameterization. Specifically, the heads of MHA were originally designed to attend to information from different representation subspaces, whereas prior studies found that some attention heads likely learn similar features and can be pruned without harming performance. Inspired by the minimum-redundancy feature selection, we assume that focusing on the most representative and distinctive features with minimum resources can mitigate the above issues and lead to more effective and efficient MHAs. I"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.14380","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/2305.14380/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":"2305.14380","created_at":"2026-07-05T07:01:01.754085+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.14380v2","created_at":"2026-07-05T07:01:01.754085+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.14380","created_at":"2026-07-05T07:01:01.754085+00:00"},{"alias_kind":"pith_short_12","alias_value":"5XWW4GZF435K","created_at":"2026-07-05T07:01:01.754085+00:00"},{"alias_kind":"pith_short_16","alias_value":"5XWW4GZF435KLLFA","created_at":"2026-07-05T07:01:01.754085+00:00"},{"alias_kind":"pith_short_8","alias_value":"5XWW4GZF","created_at":"2026-07-05T07:01:01.754085+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/5XWW4GZF435KLLFAMR5A26NVT2","json":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2.json","graph_json":"https://pith.science/api/pith-number/5XWW4GZF435KLLFAMR5A26NVT2/graph.json","events_json":"https://pith.science/api/pith-number/5XWW4GZF435KLLFAMR5A26NVT2/events.json","paper":"https://pith.science/paper/5XWW4GZF"},"agent_actions":{"view_html":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2","download_json":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2.json","view_paper":"https://pith.science/paper/5XWW4GZF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.14380&json=true","fetch_graph":"https://pith.science/api/pith-number/5XWW4GZF435KLLFAMR5A26NVT2/graph.json","fetch_events":"https://pith.science/api/pith-number/5XWW4GZF435KLLFAMR5A26NVT2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2/action/storage_attestation","attest_author":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2/action/author_attestation","sign_citation":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2/action/citation_signature","submit_replication":"https://pith.science/pith/5XWW4GZF435KLLFAMR5A26NVT2/action/replication_record"}},"created_at":"2026-07-05T07:01:01.754085+00:00","updated_at":"2026-07-05T07:01:01.754085+00:00"}