{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:U4TIVLWKYNPSZPN72QZJL5ZEOA","short_pith_number":"pith:U4TIVLWK","schema_version":"1.0","canonical_sha256":"a7268aaecac35f2cbdbfd43295f7247028b131d3466efdcb30a137c6befb11a2","source":{"kind":"arxiv","id":"2505.12250","version":1},"attestation_state":"computed","paper":{"title":"Not All Documents Are What You Need for Extracting Instruction Tuning Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chengliang Chai, Chi Zhang, Conghui He, Guoren Wang, Hongtao Li, Huaping Zhong, Jiacheng Wang, Jiantao Qiu, Jiawei Hong, Lei Cao, Tian Tan, Ye Yuan, Yizhou Yan, Yuhao Deng","submitted_at":"2025-05-18T06:10:08Z","abstract_excerpt":"Instruction tuning improves the performance of large language models (LLMs), but it heavily relies on high-quality training data. Recently, LLMs have been used to synthesize instruction data using seed question-answer (QA) pairs. However, these synthesized instructions often lack diversity and tend to be similar to the input seeds, limiting their applicability in real-world scenarios. To address this, we propose extracting instruction tuning data from web corpora that contain rich and diverse knowledge. A naive solution is to retrieve domain-specific documents and extract all QA pairs from the"},"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":"2505.12250","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-18T06:10:08Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f7820879974b2644706103cb82e3ea900594f1dd6af460145b831027b71f8872","abstract_canon_sha256":"e8538bec679037ce08a3d367de0d1bf370df7132ec1b9c42e014797cbbb5bea9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:00.895430Z","signature_b64":"VYXb1G1OPfgdWxvdUvUquImxnsEBfOLOD1aR7Zpi4pLMG8pDUCNWqMIqRRzd0SQVlbb7znmzpS5GPNRQ8gBHCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a7268aaecac35f2cbdbfd43295f7247028b131d3466efdcb30a137c6befb11a2","last_reissued_at":"2026-07-05T11:05:00.894848Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:00.894848Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Not All Documents Are What You Need for Extracting Instruction Tuning Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chengliang Chai, Chi Zhang, Conghui He, Guoren Wang, Hongtao Li, Huaping Zhong, Jiacheng Wang, Jiantao Qiu, Jiawei Hong, Lei Cao, Tian Tan, Ye Yuan, Yizhou Yan, Yuhao Deng","submitted_at":"2025-05-18T06:10:08Z","abstract_excerpt":"Instruction tuning improves the performance of large language models (LLMs), but it heavily relies on high-quality training data. Recently, LLMs have been used to synthesize instruction data using seed question-answer (QA) pairs. However, these synthesized instructions often lack diversity and tend to be similar to the input seeds, limiting their applicability in real-world scenarios. To address this, we propose extracting instruction tuning data from web corpora that contain rich and diverse knowledge. A naive solution is to retrieve domain-specific documents and extract all QA pairs from the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.12250","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/2505.12250/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":"2505.12250","created_at":"2026-07-05T11:05:00.894915+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.12250v1","created_at":"2026-07-05T11:05:00.894915+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.12250","created_at":"2026-07-05T11:05:00.894915+00:00"},{"alias_kind":"pith_short_12","alias_value":"U4TIVLWKYNPS","created_at":"2026-07-05T11:05:00.894915+00:00"},{"alias_kind":"pith_short_16","alias_value":"U4TIVLWKYNPSZPN7","created_at":"2026-07-05T11:05:00.894915+00:00"},{"alias_kind":"pith_short_8","alias_value":"U4TIVLWK","created_at":"2026-07-05T11:05:00.894915+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27931","citing_title":"DiagramRAG: A Lightweight Framework to Retrieve Scientific Diagram for Figure Generation","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA","json":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA.json","graph_json":"https://pith.science/api/pith-number/U4TIVLWKYNPSZPN72QZJL5ZEOA/graph.json","events_json":"https://pith.science/api/pith-number/U4TIVLWKYNPSZPN72QZJL5ZEOA/events.json","paper":"https://pith.science/paper/U4TIVLWK"},"agent_actions":{"view_html":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA","download_json":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA.json","view_paper":"https://pith.science/paper/U4TIVLWK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.12250&json=true","fetch_graph":"https://pith.science/api/pith-number/U4TIVLWKYNPSZPN72QZJL5ZEOA/graph.json","fetch_events":"https://pith.science/api/pith-number/U4TIVLWKYNPSZPN72QZJL5ZEOA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA/action/storage_attestation","attest_author":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA/action/author_attestation","sign_citation":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA/action/citation_signature","submit_replication":"https://pith.science/pith/U4TIVLWKYNPSZPN72QZJL5ZEOA/action/replication_record"}},"created_at":"2026-07-05T11:05:00.894915+00:00","updated_at":"2026-07-05T11:05:00.894915+00:00"}