{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XOSYVZSILQ75E6VF3BCQ43XWDH","short_pith_number":"pith:XOSYVZSI","schema_version":"1.0","canonical_sha256":"bba58ae6485c3fd27aa5d8450e6ef619c4e7f5ce0579d097628adcb63c27ad4c","source":{"kind":"arxiv","id":"2602.01572","version":2},"attestation_state":"computed","paper":{"title":"LLM-based Embeddings: Attention Values Encode Sentence Semantics Better Than Hidden States","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Cam-Tu Nguyen, Jiaxuan Chen, Ke Qin, Yeqin Zhang, Yizheng Zhao, Yunfei Wang","submitted_at":"2026-02-02T03:09:37Z","abstract_excerpt":"Sentence representations are foundational to many Natural Language Processing (NLP) applications. While recent methods leverage Large Language Models (LLMs) to derive sentence representations, most rely on final-layer hidden states, which are optimized for next-token prediction and thus often fail to capture global, sentence-level semantics. This paper introduces a novel perspective, demonstrating that attention value vectors capture sentence semantics more effectively than hidden states. We propose Value Aggregation (VA), a simple method that pools token values across multiple layers and toke"},"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":"2602.01572","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-02T03:09:37Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"13ad9659a4b467570e97cf12019107e954a00c53d87bab821596b465098e0c67","abstract_canon_sha256":"dc8735bdd9878ced3d2782bf44cc7ccbc49338a9e85bfb93892e8c1a8bad37d4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-12T01:09:21.558676Z","signature_b64":"zPYwOZMxbnh9UVF/0AlhSfAzx90Ut3mGmJLMZXiKSuJLB8nUdFAY/ikcVWNi96nZ6UXBBy/becO8JAiUNgOZBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bba58ae6485c3fd27aa5d8450e6ef619c4e7f5ce0579d097628adcb63c27ad4c","last_reissued_at":"2026-06-12T01:09:21.557805Z","signature_status":"signed_v1","first_computed_at":"2026-06-12T01:09:21.557805Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM-based Embeddings: Attention Values Encode Sentence Semantics Better Than Hidden States","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Cam-Tu Nguyen, Jiaxuan Chen, Ke Qin, Yeqin Zhang, Yizheng Zhao, Yunfei Wang","submitted_at":"2026-02-02T03:09:37Z","abstract_excerpt":"Sentence representations are foundational to many Natural Language Processing (NLP) applications. While recent methods leverage Large Language Models (LLMs) to derive sentence representations, most rely on final-layer hidden states, which are optimized for next-token prediction and thus often fail to capture global, sentence-level semantics. This paper introduces a novel perspective, demonstrating that attention value vectors capture sentence semantics more effectively than hidden states. We propose Value Aggregation (VA), a simple method that pools token values across multiple layers and toke"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.01572","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/2602.01572/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":"2602.01572","created_at":"2026-06-12T01:09:21.557908+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.01572v2","created_at":"2026-06-12T01:09:21.557908+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.01572","created_at":"2026-06-12T01:09:21.557908+00:00"},{"alias_kind":"pith_short_12","alias_value":"XOSYVZSILQ75","created_at":"2026-06-12T01:09:21.557908+00:00"},{"alias_kind":"pith_short_16","alias_value":"XOSYVZSILQ75E6VF","created_at":"2026-06-12T01:09:21.557908+00:00"},{"alias_kind":"pith_short_8","alias_value":"XOSYVZSI","created_at":"2026-06-12T01:09:21.557908+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/XOSYVZSILQ75E6VF3BCQ43XWDH","json":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH.json","graph_json":"https://pith.science/api/pith-number/XOSYVZSILQ75E6VF3BCQ43XWDH/graph.json","events_json":"https://pith.science/api/pith-number/XOSYVZSILQ75E6VF3BCQ43XWDH/events.json","paper":"https://pith.science/paper/XOSYVZSI"},"agent_actions":{"view_html":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH","download_json":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH.json","view_paper":"https://pith.science/paper/XOSYVZSI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.01572&json=true","fetch_graph":"https://pith.science/api/pith-number/XOSYVZSILQ75E6VF3BCQ43XWDH/graph.json","fetch_events":"https://pith.science/api/pith-number/XOSYVZSILQ75E6VF3BCQ43XWDH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH/action/storage_attestation","attest_author":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH/action/author_attestation","sign_citation":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH/action/citation_signature","submit_replication":"https://pith.science/pith/XOSYVZSILQ75E6VF3BCQ43XWDH/action/replication_record"}},"created_at":"2026-06-12T01:09:21.557908+00:00","updated_at":"2026-06-12T01:09:21.557908+00:00"}