{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UIDWNJCRD5HTYRPK3EJM4IA5YC","short_pith_number":"pith:UIDWNJCR","schema_version":"1.0","canonical_sha256":"a20766a4511f4f3c45ead912ce201dc0abbad14d0f7189b3dfc8dd64cc40438b","source":{"kind":"arxiv","id":"2409.02727","version":2},"attestation_state":"computed","paper":{"title":"Pooling And Attention: What Are Effective Designs For LLM-Based Embedding Models?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Yixuan Tang, Yi Yang","submitted_at":"2024-09-04T14:01:48Z","abstract_excerpt":"The significant advancements of Large Language Models (LLMs) in generative tasks have led to a growing body of work exploring LLM-based embedding models. While these models, employing different pooling and attention strategies, have achieved state-of-the-art performance on public embedding benchmarks, questions still arise about what constitutes an effective design for LLM-based embedding models. However, these models are often trained on different datasets, using different LLM base models or training settings. Moreover, evaluations on public embedding benchmarks often fail to report statistic"},"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":"2409.02727","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-04T14:01:48Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"b623c783cbea8f5a308bd4de077044db1e87323d22a02bed2433d41561ac25e4","abstract_canon_sha256":"160f98bd6447f4386ea40f8ce7bb8fcfd9bf566bf00118207e23e7c5b97b0506"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:03:23.761000Z","signature_b64":"uYnytdy5SUNK40dPTjsBZVC8TEsTmHwtzCby9ubDVWvAmadUu87KvJOdv1kplDvJmAc5f9kcb7Jp3iLnaI1MAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a20766a4511f4f3c45ead912ce201dc0abbad14d0f7189b3dfc8dd64cc40438b","last_reissued_at":"2026-07-05T09:03:23.760478Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:03:23.760478Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pooling And Attention: What Are Effective Designs For LLM-Based Embedding Models?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CL","authors_text":"Yixuan Tang, Yi Yang","submitted_at":"2024-09-04T14:01:48Z","abstract_excerpt":"The significant advancements of Large Language Models (LLMs) in generative tasks have led to a growing body of work exploring LLM-based embedding models. While these models, employing different pooling and attention strategies, have achieved state-of-the-art performance on public embedding benchmarks, questions still arise about what constitutes an effective design for LLM-based embedding models. However, these models are often trained on different datasets, using different LLM base models or training settings. Moreover, evaluations on public embedding benchmarks often fail to report statistic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.02727","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/2409.02727/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":"2409.02727","created_at":"2026-07-05T09:03:23.760543+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.02727v2","created_at":"2026-07-05T09:03:23.760543+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.02727","created_at":"2026-07-05T09:03:23.760543+00:00"},{"alias_kind":"pith_short_12","alias_value":"UIDWNJCRD5HT","created_at":"2026-07-05T09:03:23.760543+00:00"},{"alias_kind":"pith_short_16","alias_value":"UIDWNJCRD5HTYRPK","created_at":"2026-07-05T09:03:23.760543+00:00"},{"alias_kind":"pith_short_8","alias_value":"UIDWNJCR","created_at":"2026-07-05T09:03:23.760543+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11023","citing_title":"Generative Archetype-Grounded Item Representations for Sequential Recommendation","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08191","citing_title":"Frequency-Domain Latent Attention Gating for Cross-Domain Token Aggregation","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08260","citing_title":"Behavior-Aware Item Modeling via Dynamic Procedural Solution Representations for Knowledge Tracing","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07775","citing_title":"POETS: Uncertainty-Aware LLM Optimization via Compute-Efficient Policy Ensembles","ref_index":111,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18519","citing_title":"LLM Safety From Within: Detecting Harmful Content with Internal Representations","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25605","citing_title":"Health System Scale Semantic Search Across Unstructured Clinical Notes","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC","json":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC.json","graph_json":"https://pith.science/api/pith-number/UIDWNJCRD5HTYRPK3EJM4IA5YC/graph.json","events_json":"https://pith.science/api/pith-number/UIDWNJCRD5HTYRPK3EJM4IA5YC/events.json","paper":"https://pith.science/paper/UIDWNJCR"},"agent_actions":{"view_html":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC","download_json":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC.json","view_paper":"https://pith.science/paper/UIDWNJCR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.02727&json=true","fetch_graph":"https://pith.science/api/pith-number/UIDWNJCRD5HTYRPK3EJM4IA5YC/graph.json","fetch_events":"https://pith.science/api/pith-number/UIDWNJCRD5HTYRPK3EJM4IA5YC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC/action/storage_attestation","attest_author":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC/action/author_attestation","sign_citation":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC/action/citation_signature","submit_replication":"https://pith.science/pith/UIDWNJCRD5HTYRPK3EJM4IA5YC/action/replication_record"}},"created_at":"2026-07-05T09:03:23.760543+00:00","updated_at":"2026-07-05T09:03:23.760543+00:00"}