{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BVNT5NGGPYUHWGBJTBBIZ4SY2Z","short_pith_number":"pith:BVNT5NGG","schema_version":"1.0","canonical_sha256":"0d5b3eb4c67e287b182998428cf258d673facf83f2c02149f22e193f838b2209","source":{"kind":"arxiv","id":"2605.30348","version":1},"attestation_state":"computed","paper":{"title":"LLMSurgeon: Diagnosing Data Mixture of Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jiacheng Cui, Jiacheng Liu, Xiaohan Zhao, Xinyi Shang, Xinyue Bi, Yaxin Luo, Zhaoyi Li, Zhiqiang Shen","submitted_at":"2026-05-28T17:59:53Z","abstract_excerpt":"The pretraining data mixture of Large Language Models (LLMs) constitutes their \"digital DNA\", shaping model behaviors, capabilities, and failure modes. Yet this composition is rarely disclosed, making post-hoc auditing of data combination or provenance difficult. In this work, we formalize $\\textbf{{Data Mixture Surgery (DMS)}}$: given only generated text from a target LLM, estimate the domain-level distribution of its pretraining corpus under a predefined taxonomy. We propose $\\textbf{{LLMSurgeon}}$, a strong framework that casts DMS as an inverse problem under the label-shift assumption. Rat"},"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":"2605.30348","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-28T17:59:53Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"837bdfddf728c638869dadc200c7be58ea6f18b1896d29698fad39a83775458b","abstract_canon_sha256":"93400ac8a380356bc3f54ce5bd807f559f1adc87f6006aa2b84099d086ed64e9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:06:17.281219Z","signature_b64":"2D1aBe/U5G0tvesY6FQFPO3iyk2yEPA03VIJRS4oiYEX4uA/NysL4qSpKVbglueuuowBDjzRtf67uhr/v6I0Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0d5b3eb4c67e287b182998428cf258d673facf83f2c02149f22e193f838b2209","last_reissued_at":"2026-05-29T02:06:17.280792Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:06:17.280792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLMSurgeon: Diagnosing Data Mixture of Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jiacheng Cui, Jiacheng Liu, Xiaohan Zhao, Xinyi Shang, Xinyue Bi, Yaxin Luo, Zhaoyi Li, Zhiqiang Shen","submitted_at":"2026-05-28T17:59:53Z","abstract_excerpt":"The pretraining data mixture of Large Language Models (LLMs) constitutes their \"digital DNA\", shaping model behaviors, capabilities, and failure modes. Yet this composition is rarely disclosed, making post-hoc auditing of data combination or provenance difficult. In this work, we formalize $\\textbf{{Data Mixture Surgery (DMS)}}$: given only generated text from a target LLM, estimate the domain-level distribution of its pretraining corpus under a predefined taxonomy. We propose $\\textbf{{LLMSurgeon}}$, a strong framework that casts DMS as an inverse problem under the label-shift assumption. Rat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.30348","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/2605.30348/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":"2605.30348","created_at":"2026-05-29T02:06:17.280856+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.30348v1","created_at":"2026-05-29T02:06:17.280856+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.30348","created_at":"2026-05-29T02:06:17.280856+00:00"},{"alias_kind":"pith_short_12","alias_value":"BVNT5NGGPYUH","created_at":"2026-05-29T02:06:17.280856+00:00"},{"alias_kind":"pith_short_16","alias_value":"BVNT5NGGPYUHWGBJ","created_at":"2026-05-29T02:06:17.280856+00:00"},{"alias_kind":"pith_short_8","alias_value":"BVNT5NGG","created_at":"2026-05-29T02:06:17.280856+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z","json":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z.json","graph_json":"https://pith.science/api/pith-number/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/graph.json","events_json":"https://pith.science/api/pith-number/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/events.json","paper":"https://pith.science/paper/BVNT5NGG"},"agent_actions":{"view_html":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z","download_json":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z.json","view_paper":"https://pith.science/paper/BVNT5NGG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.30348&json=true","fetch_graph":"https://pith.science/api/pith-number/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/graph.json","fetch_events":"https://pith.science/api/pith-number/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/action/storage_attestation","attest_author":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/action/author_attestation","sign_citation":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/action/citation_signature","submit_replication":"https://pith.science/pith/BVNT5NGGPYUHWGBJTBBIZ4SY2Z/action/replication_record"}},"created_at":"2026-05-29T02:06:17.280856+00:00","updated_at":"2026-05-29T02:06:17.280856+00:00"}