{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:AHSLECIHUKA77BSFGYTQWTS3LQ","short_pith_number":"pith:AHSLECIH","schema_version":"1.0","canonical_sha256":"01e4b20907a281ff864536270b4e5b5c2a4daea7a46d15beebdc39f95bfc02f2","source":{"kind":"arxiv","id":"2505.20416","version":1},"attestation_state":"computed","paper":{"title":"GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Huanjun Kong, Jinzhe Li, Nanqing Dong, Wanli Jiang, Wanli Ouyang, Zhonghang Yuan, Zihong Chen","submitted_at":"2025-05-26T18:06:50Z","abstract_excerpt":"Fine-tuning for large language models (LLMs) typically requires substantial amounts of high-quality supervised data, which is both costly and labor-intensive to acquire. While synthetic data generation has emerged as a promising solution, existing approaches frequently suffer from factual inaccuracies, insufficient long-tail coverage, simplistic knowledge structures, and homogenized outputs. To address these challenges, we introduce GraphGen, a knowledge graph-guided framework designed for three key question-answering (QA) scenarios: atomic QA, aggregated QA, and multi-hop QA. It begins by con"},"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":"2505.20416","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-26T18:06:50Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c019b3514dbd9b854023d381904bc0c525ead44fdee008cf81321dfcf08c2322","abstract_canon_sha256":"b31368a2d96d82f3e5ecdbf2b237b2984c161c1c0e7245ac36b29e2bee299390"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:12.298001Z","signature_b64":"urAB0S5Gi89bfZIqmxAYM36Ngxy5R9jvkyq2ZPqNNzs7gt839vWfAItNjOpzbyAHElbZKcIpuHGg8/5U2O6HBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"01e4b20907a281ff864536270b4e5b5c2a4daea7a46d15beebdc39f95bfc02f2","last_reissued_at":"2026-07-05T11:10:12.297537Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:12.297537Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Huanjun Kong, Jinzhe Li, Nanqing Dong, Wanli Jiang, Wanli Ouyang, Zhonghang Yuan, Zihong Chen","submitted_at":"2025-05-26T18:06:50Z","abstract_excerpt":"Fine-tuning for large language models (LLMs) typically requires substantial amounts of high-quality supervised data, which is both costly and labor-intensive to acquire. While synthetic data generation has emerged as a promising solution, existing approaches frequently suffer from factual inaccuracies, insufficient long-tail coverage, simplistic knowledge structures, and homogenized outputs. To address these challenges, we introduce GraphGen, a knowledge graph-guided framework designed for three key question-answering (QA) scenarios: atomic QA, aggregated QA, and multi-hop QA. It begins by con"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.20416","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/2505.20416/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":"2505.20416","created_at":"2026-07-05T11:10:12.297588+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.20416v1","created_at":"2026-07-05T11:10:12.297588+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.20416","created_at":"2026-07-05T11:10:12.297588+00:00"},{"alias_kind":"pith_short_12","alias_value":"AHSLECIHUKA7","created_at":"2026-07-05T11:10:12.297588+00:00"},{"alias_kind":"pith_short_16","alias_value":"AHSLECIHUKA77BSF","created_at":"2026-07-05T11:10:12.297588+00:00"},{"alias_kind":"pith_short_8","alias_value":"AHSLECIH","created_at":"2026-07-05T11:10:12.297588+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12837","citing_title":"LoHoSearch: Benchmarking Long-Horizon Search Agents Beyond the Human Difficulty Ceiling","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12087","citing_title":"FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18025","citing_title":"TeleCom-Bench: How Far Are Large Language Models from Industrial Telecommunication Applications?","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19394","citing_title":"EmbGen: Teaching with Reassembled Corpora","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2602.12705","citing_title":"MedXIAOHE: A Comprehensive Recipe for Building Medical MLLMs","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08709","citing_title":"UniShield: Unified Face Attack Detection via KG-Informed Multimodal Reasoning","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ","json":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ.json","graph_json":"https://pith.science/api/pith-number/AHSLECIHUKA77BSFGYTQWTS3LQ/graph.json","events_json":"https://pith.science/api/pith-number/AHSLECIHUKA77BSFGYTQWTS3LQ/events.json","paper":"https://pith.science/paper/AHSLECIH"},"agent_actions":{"view_html":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ","download_json":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ.json","view_paper":"https://pith.science/paper/AHSLECIH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.20416&json=true","fetch_graph":"https://pith.science/api/pith-number/AHSLECIHUKA77BSFGYTQWTS3LQ/graph.json","fetch_events":"https://pith.science/api/pith-number/AHSLECIHUKA77BSFGYTQWTS3LQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ/action/storage_attestation","attest_author":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ/action/author_attestation","sign_citation":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ/action/citation_signature","submit_replication":"https://pith.science/pith/AHSLECIHUKA77BSFGYTQWTS3LQ/action/replication_record"}},"created_at":"2026-07-05T11:10:12.297588+00:00","updated_at":"2026-07-05T11:10:12.297588+00:00"}