{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:77PPNT5YOWT3WUHHUXBW2RQPRP","short_pith_number":"pith:77PPNT5Y","schema_version":"1.0","canonical_sha256":"ffdef6cfb875a7bb50e7a5c36d460f8bfbd66da8f7761740a40ddb72fe444727","source":{"kind":"arxiv","id":"2502.15526","version":1},"attestation_state":"computed","paper":{"title":"Scaling Sparse and Dense Retrieval in Decoder-Only LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Hamed Zamani, Hansi Zeng, Julian Killingback","submitted_at":"2025-02-21T15:28:26Z","abstract_excerpt":"Scaling large language models (LLMs) has shown great potential for improving retrieval model performance; however, previous studies have mainly focused on dense retrieval trained with contrastive loss (CL), neglecting the scaling behavior of other retrieval paradigms and optimization techniques, such as sparse retrieval and knowledge distillation (KD). In this work, we conduct a systematic comparative study on how different retrieval paradigms (sparse vs. dense) and fine-tuning objectives (CL vs. KD vs. their combination) affect retrieval performance across different model scales. Using MSMARC"},"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":"2502.15526","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2025-02-21T15:28:26Z","cross_cats_sorted":[],"title_canon_sha256":"68e0b0da9f72522b0c2f224bc95c91ae335fbc831a7c75be3f789eb8285e9c99","abstract_canon_sha256":"48e490bb56372eccd7c6914be55808578a45689cef6453b729aa55ee511ea92a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:03.010280Z","signature_b64":"us4VT4uWFFZStMF10hTfhBOsK8GKG8JVqZMRu9bGvIjMe4+DI4mLX49wpvi+x3T2zJrBtf+AmRyh2dod4GJQDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ffdef6cfb875a7bb50e7a5c36d460f8bfbd66da8f7761740a40ddb72fe444727","last_reissued_at":"2026-07-05T10:18:03.009803Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:03.009803Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Sparse and Dense Retrieval in Decoder-Only LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Hamed Zamani, Hansi Zeng, Julian Killingback","submitted_at":"2025-02-21T15:28:26Z","abstract_excerpt":"Scaling large language models (LLMs) has shown great potential for improving retrieval model performance; however, previous studies have mainly focused on dense retrieval trained with contrastive loss (CL), neglecting the scaling behavior of other retrieval paradigms and optimization techniques, such as sparse retrieval and knowledge distillation (KD). In this work, we conduct a systematic comparative study on how different retrieval paradigms (sparse vs. dense) and fine-tuning objectives (CL vs. KD vs. their combination) affect retrieval performance across different model scales. Using MSMARC"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15526","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/2502.15526/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":"2502.15526","created_at":"2026-07-05T10:18:03.009902+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.15526v1","created_at":"2026-07-05T10:18:03.009902+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15526","created_at":"2026-07-05T10:18:03.009902+00:00"},{"alias_kind":"pith_short_12","alias_value":"77PPNT5YOWT3","created_at":"2026-07-05T10:18:03.009902+00:00"},{"alias_kind":"pith_short_16","alias_value":"77PPNT5YOWT3WUHH","created_at":"2026-07-05T10:18:03.009902+00:00"},{"alias_kind":"pith_short_8","alias_value":"77PPNT5Y","created_at":"2026-07-05T10:18:03.009902+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.04734","citing_title":"Beyond Hard Negatives: The Importance of Score Distribution in Knowledge Distillation for Dense Retrieval","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP","json":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP.json","graph_json":"https://pith.science/api/pith-number/77PPNT5YOWT3WUHHUXBW2RQPRP/graph.json","events_json":"https://pith.science/api/pith-number/77PPNT5YOWT3WUHHUXBW2RQPRP/events.json","paper":"https://pith.science/paper/77PPNT5Y"},"agent_actions":{"view_html":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP","download_json":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP.json","view_paper":"https://pith.science/paper/77PPNT5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.15526&json=true","fetch_graph":"https://pith.science/api/pith-number/77PPNT5YOWT3WUHHUXBW2RQPRP/graph.json","fetch_events":"https://pith.science/api/pith-number/77PPNT5YOWT3WUHHUXBW2RQPRP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP/action/storage_attestation","attest_author":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP/action/author_attestation","sign_citation":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP/action/citation_signature","submit_replication":"https://pith.science/pith/77PPNT5YOWT3WUHHUXBW2RQPRP/action/replication_record"}},"created_at":"2026-07-05T10:18:03.009902+00:00","updated_at":"2026-07-05T10:18:03.009902+00:00"}