{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:B5WOMRVASFLFUUS4VBNSSQJ2HR","short_pith_number":"pith:B5WOMRVA","schema_version":"1.0","canonical_sha256":"0f6ce646a091565a525ca85b29413a3c69829219b01220394513e136411f94e5","source":{"kind":"arxiv","id":"2606.04929","version":1},"attestation_state":"computed","paper":{"title":"Sequential Data Poisoning in LLM Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Gautam Kamath, Jack Sanderson, Xiaoqian Lu, Yihan Wang, Yiwei Lu","submitted_at":"2026-06-03T14:22:50Z","abstract_excerpt":"LLM post-training proceeds through multiple stages, e.g., supervised fine-tuning (SFT) followed by reinforcement learning from human feedback (RLHF) or direct preference optimization (DPO), where each stage draws data from different, potentially untrusted sources. Existing literature assumes data poisoning attacks may occur at each training stage, but neglects the possibility of multiple attackers. To study the trustworthiness of the entire post-training pipeline, we propose the threat model of sequential data poisoning, where multiple adversaries separately poison the SFT and preference datas"},"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":"2606.04929","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T14:22:50Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"b12462f880fb8188711d4cc965d452928a7fef883c28027c6f304d2256943863","abstract_canon_sha256":"63745c8ae723f70fabb3d4a6652adda8b1d7ceaab80d1e883543ee79846c8bb0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:09:56.223278Z","signature_b64":"584YcAzPgigf9HMq+P/2ZAEN6laxrMKbXLm0Pk3N+rQioi+2TBIi1ZR9j+RVKsVLttFYJ0FR+dps53wLXxAnCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f6ce646a091565a525ca85b29413a3c69829219b01220394513e136411f94e5","last_reissued_at":"2026-06-04T01:09:56.222508Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:09:56.222508Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sequential Data Poisoning in LLM Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Gautam Kamath, Jack Sanderson, Xiaoqian Lu, Yihan Wang, Yiwei Lu","submitted_at":"2026-06-03T14:22:50Z","abstract_excerpt":"LLM post-training proceeds through multiple stages, e.g., supervised fine-tuning (SFT) followed by reinforcement learning from human feedback (RLHF) or direct preference optimization (DPO), where each stage draws data from different, potentially untrusted sources. Existing literature assumes data poisoning attacks may occur at each training stage, but neglects the possibility of multiple attackers. To study the trustworthiness of the entire post-training pipeline, we propose the threat model of sequential data poisoning, where multiple adversaries separately poison the SFT and preference datas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04929","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/2606.04929/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":"2606.04929","created_at":"2026-06-04T01:09:56.222636+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.04929v1","created_at":"2026-06-04T01:09:56.222636+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04929","created_at":"2026-06-04T01:09:56.222636+00:00"},{"alias_kind":"pith_short_12","alias_value":"B5WOMRVASFLF","created_at":"2026-06-04T01:09:56.222636+00:00"},{"alias_kind":"pith_short_16","alias_value":"B5WOMRVASFLFUUS4","created_at":"2026-06-04T01:09:56.222636+00:00"},{"alias_kind":"pith_short_8","alias_value":"B5WOMRVA","created_at":"2026-06-04T01:09:56.222636+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/B5WOMRVASFLFUUS4VBNSSQJ2HR","json":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR.json","graph_json":"https://pith.science/api/pith-number/B5WOMRVASFLFUUS4VBNSSQJ2HR/graph.json","events_json":"https://pith.science/api/pith-number/B5WOMRVASFLFUUS4VBNSSQJ2HR/events.json","paper":"https://pith.science/paper/B5WOMRVA"},"agent_actions":{"view_html":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR","download_json":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR.json","view_paper":"https://pith.science/paper/B5WOMRVA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.04929&json=true","fetch_graph":"https://pith.science/api/pith-number/B5WOMRVASFLFUUS4VBNSSQJ2HR/graph.json","fetch_events":"https://pith.science/api/pith-number/B5WOMRVASFLFUUS4VBNSSQJ2HR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR/action/storage_attestation","attest_author":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR/action/author_attestation","sign_citation":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR/action/citation_signature","submit_replication":"https://pith.science/pith/B5WOMRVASFLFUUS4VBNSSQJ2HR/action/replication_record"}},"created_at":"2026-06-04T01:09:56.222636+00:00","updated_at":"2026-06-04T01:09:56.222636+00:00"}