{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SHCITBJ4IFQC4PDSF5GCCWUJRI","short_pith_number":"pith:SHCITBJ4","schema_version":"1.0","canonical_sha256":"91c489853c41602e3c722f4c215a898a0ca0d7aaea3aa5bcae22185236186d39","source":{"kind":"arxiv","id":"2509.02896","version":2},"attestation_state":"computed","paper":{"title":"Cut Costs, Not Accuracy: LLM-Powered Data Processing with Guarantees","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DB","authors_text":"Aditya Parameswaran, Sepanta Zeighami, Shreya Shankar","submitted_at":"2025-09-02T23:41:50Z","abstract_excerpt":"Large Language Models (LLMs) are being increasingly used as a building block in data systems to process large text datasets. To do so, LLM model providers offer multiple LLMs with different sizes, spanning various cost-quality trade-offs when processing text at scale. Top-of-the-line LLMs (e.g., GPT-4o, Claude Sonnet) operate with high accuracy but are prohibitively expensive when processing many records. To avoid high costs, more affordable but lower quality LLMs (e.g., GPT-4o-mini, Claude Haiku) can be used to process records, but we need to ensure that the overall accuracy does not deviate "},"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":"2509.02896","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DB","submitted_at":"2025-09-02T23:41:50Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"21402d1d1aeaa1ea50340b55c89adad7439d0f50079aa9b6046ea3994ae58bad","abstract_canon_sha256":"77e38be9d99dce9be87679e75a280d7de1baefdff89559b513629454e0f16eb9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:10:59.955873Z","signature_b64":"8XQGm3JbzeFI+I+i5CfDiWPunInfD0Mk65adVM7vvPiB4S0rkxTD48k+DlVGiY4wN2iyaLhvKwgqKfbDwvdlAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"91c489853c41602e3c722f4c215a898a0ca0d7aaea3aa5bcae22185236186d39","last_reissued_at":"2026-07-05T12:10:59.955343Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:10:59.955343Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cut Costs, Not Accuracy: LLM-Powered Data Processing with Guarantees","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DB","authors_text":"Aditya Parameswaran, Sepanta Zeighami, Shreya Shankar","submitted_at":"2025-09-02T23:41:50Z","abstract_excerpt":"Large Language Models (LLMs) are being increasingly used as a building block in data systems to process large text datasets. To do so, LLM model providers offer multiple LLMs with different sizes, spanning various cost-quality trade-offs when processing text at scale. Top-of-the-line LLMs (e.g., GPT-4o, Claude Sonnet) operate with high accuracy but are prohibitively expensive when processing many records. To avoid high costs, more affordable but lower quality LLMs (e.g., GPT-4o-mini, Claude Haiku) can be used to process records, but we need to ensure that the overall accuracy does not deviate "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.02896","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/2509.02896/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":"2509.02896","created_at":"2026-07-05T12:10:59.955402+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.02896v2","created_at":"2026-07-05T12:10:59.955402+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.02896","created_at":"2026-07-05T12:10:59.955402+00:00"},{"alias_kind":"pith_short_12","alias_value":"SHCITBJ4IFQC","created_at":"2026-07-05T12:10:59.955402+00:00"},{"alias_kind":"pith_short_16","alias_value":"SHCITBJ4IFQC4PDS","created_at":"2026-07-05T12:10:59.955402+00:00"},{"alias_kind":"pith_short_8","alias_value":"SHCITBJ4","created_at":"2026-07-05T12:10:59.955402+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08090","citing_title":"Fast LLM-Based Semantic Filtering: From a Unified Framework to an Adaptive Two-Phase Method","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI","json":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI.json","graph_json":"https://pith.science/api/pith-number/SHCITBJ4IFQC4PDSF5GCCWUJRI/graph.json","events_json":"https://pith.science/api/pith-number/SHCITBJ4IFQC4PDSF5GCCWUJRI/events.json","paper":"https://pith.science/paper/SHCITBJ4"},"agent_actions":{"view_html":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI","download_json":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI.json","view_paper":"https://pith.science/paper/SHCITBJ4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.02896&json=true","fetch_graph":"https://pith.science/api/pith-number/SHCITBJ4IFQC4PDSF5GCCWUJRI/graph.json","fetch_events":"https://pith.science/api/pith-number/SHCITBJ4IFQC4PDSF5GCCWUJRI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI/action/storage_attestation","attest_author":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI/action/author_attestation","sign_citation":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI/action/citation_signature","submit_replication":"https://pith.science/pith/SHCITBJ4IFQC4PDSF5GCCWUJRI/action/replication_record"}},"created_at":"2026-07-05T12:10:59.955402+00:00","updated_at":"2026-07-05T12:10:59.955402+00:00"}