{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:TA5KJQNT3FHIQUNJIMRGMI64ZB","short_pith_number":"pith:TA5KJQNT","canonical_record":{"source":{"id":"2605.02495","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T11:45:38Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"6ce72836f72c2d4bc21da7421672596e800250bd45781c68f4170bcb702393f2","abstract_canon_sha256":"168897d1319adf9cf5ca6ec100e90635defd2020f0090cbb82ced51119b27a71"},"schema_version":"1.0"},"canonical_sha256":"983aa4c1b3d94e8851a943226623dcc84bdf1b058985d897d42c7897ef4bb2de","source":{"kind":"arxiv","id":"2605.02495","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.02495","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.02495v2","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.02495","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_12","alias_value":"TA5KJQNT3FHI","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_16","alias_value":"TA5KJQNT3FHIQUNJ","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_8","alias_value":"TA5KJQNT","created_at":"2026-05-26T02:04:11Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:TA5KJQNT3FHIQUNJIMRGMI64ZB","target":"record","payload":{"canonical_record":{"source":{"id":"2605.02495","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T11:45:38Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"6ce72836f72c2d4bc21da7421672596e800250bd45781c68f4170bcb702393f2","abstract_canon_sha256":"168897d1319adf9cf5ca6ec100e90635defd2020f0090cbb82ced51119b27a71"},"schema_version":"1.0"},"canonical_sha256":"983aa4c1b3d94e8851a943226623dcc84bdf1b058985d897d42c7897ef4bb2de","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:11.842217Z","signature_b64":"ObyHYb7YYsiGUjDOEZPq1Vi0G7mf/VBN5ncMHv0S444cFv7MumSQD9zGaZ9ZA5Aj4+hI8vVsUaCNIvp+jnllBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"983aa4c1b3d94e8851a943226623dcc84bdf1b058985d897d42c7897ef4bb2de","last_reissued_at":"2026-05-26T02:04:11.841288Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:11.841288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.02495","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T02:04:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HNUFFsYwDXeXW1pfI7TX+IJ8xhrwI4zdyczWR+Q8UNx7+A2hrJ9ZZMdQWkeeEU5KLTr7Qhi1ofYFYoJoMdZ6CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T02:13:50.492133Z"},"content_sha256":"12fa24d1fe832370556d371a27b5623014c63073ddfe9bf4f045f039bfd87fe5","schema_version":"1.0","event_id":"sha256:12fa24d1fe832370556d371a27b5623014c63073ddfe9bf4f045f039bfd87fe5"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:TA5KJQNT3FHIQUNJIMRGMI64ZB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Efficient Preference Poisoning Attack on Offline RLHF","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem.","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chenye Yang, Lifeng Lai, Weiyu Xu","submitted_at":"2026-05-04T11:45:38Z","abstract_excerpt":"Offline Reinforcement Learning from Human Feedback (RLHF) pipelines such as Direct Preference Optimization (DPO) train on a pre-collected preference dataset, which makes them vulnerable to preference poisoning attack. We study label flip attacks against log-linear DPO. We first illustrate that flipping one preference label induces a parameter-independent shift in the DPO gradient. Using this key property, we can then convert the targeted poisoning problem into a structured binary sparse approximation problem. To solve this problem, we develop two attack methods: Binary-Aware Lattice Attack (BA"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Flipping one preference label induces a parameter-independent shift in the DPO gradient, converting the targeted poisoning problem into a structured binary sparse approximation problem that BAL-A and BMP-A solve with sufficient conditions and coherence-based guarantees.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The central reduction assumes that the DPO objective is strictly log-linear in the parameters so that the gradient shift from a single label flip remains independent of the current parameter vector; if this independence fails for realistic non-linear models or regularizers, the conversion to sparse approximation no longer holds.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Label-flip attacks on log-linear DPO reduce to binary sparse approximation problems that can be solved efficiently by lattice-based and binary matching pursuit methods with recovery guarantees.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0fc296242b5d6b9b57bfa821b2bdaffbc309c1b3095021e6606ad1e709894733"},"source":{"id":"2605.02495","kind":"arxiv","version":2},"verdict":{"id":"4c853f26-0168-484b-a44f-1099298b1503","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T19:25:48.070349Z","strongest_claim":"Flipping one preference label induces a parameter-independent shift in the DPO gradient, converting the targeted poisoning problem into a structured binary sparse approximation problem that BAL-A and BMP-A solve with sufficient conditions and coherence-based guarantees.","one_line_summary":"Label-flip attacks on log-linear DPO reduce to binary sparse approximation problems that can be solved efficiently by lattice-based and binary matching pursuit methods with recovery guarantees.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The central reduction assumes that the DPO objective is strictly log-linear in the parameters so that the gradient shift from a single label flip remains independent of the current parameter vector; if this independence fails for realistic non-linear models or regularizers, the conversion to sparse approximation no longer holds.","pith_extraction_headline":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.02495/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T15:39:43.237812Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T03:01:22.782107Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:20:23.175153Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"e123e53f1d28080b036ab00e272a95fc00fabdab7e773726f4130eb401745d9b"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":3,"snapshot_sha256":"3cab39d4d8f1f232287399b979029d5e29bbc975550da0d3ebcd421fd2e15205"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"4c853f26-0168-484b-a44f-1099298b1503"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T02:04:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9+Z3iOipDPVxHWQT/AYS2qkzHPokgIBmuQwuG9oCsMyHERr8NJsrLOGGOMEgrX5hN2P9xuuseMI3ctIHPK9BAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T02:13:50.493053Z"},"content_sha256":"9f45ddf865fd1f2a21fc66f51b6f6203d3ce0884feb01ddef990eb47159406c9","schema_version":"1.0","event_id":"sha256:9f45ddf865fd1f2a21fc66f51b6f6203d3ce0884feb01ddef990eb47159406c9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/bundle.json","state_url":"https://pith.science/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-31T02:13:50Z","links":{"resolver":"https://pith.science/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB","bundle":"https://pith.science/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/bundle.json","state":"https://pith.science/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/TA5KJQNT3FHIQUNJIMRGMI64ZB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:TA5KJQNT3FHIQUNJIMRGMI64ZB","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"168897d1319adf9cf5ca6ec100e90635defd2020f0090cbb82ced51119b27a71","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T11:45:38Z","title_canon_sha256":"6ce72836f72c2d4bc21da7421672596e800250bd45781c68f4170bcb702393f2"},"schema_version":"1.0","source":{"id":"2605.02495","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.02495","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.02495v2","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.02495","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_12","alias_value":"TA5KJQNT3FHI","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_16","alias_value":"TA5KJQNT3FHIQUNJ","created_at":"2026-05-26T02:04:11Z"},{"alias_kind":"pith_short_8","alias_value":"TA5KJQNT","created_at":"2026-05-26T02:04:11Z"}],"graph_snapshots":[{"event_id":"sha256:9f45ddf865fd1f2a21fc66f51b6f6203d3ce0884feb01ddef990eb47159406c9","target":"graph","created_at":"2026-05-26T02:04:11Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Flipping one preference label induces a parameter-independent shift in the DPO gradient, converting the targeted poisoning problem into a structured binary sparse approximation problem that BAL-A and BMP-A solve with sufficient conditions and coherence-based guarantees."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The central reduction assumes that the DPO objective is strictly log-linear in the parameters so that the gradient shift from a single label flip remains independent of the current parameter vector; if this independence fails for realistic non-linear models or regularizers, the conversion to sparse approximation no longer holds."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Label-flip attacks on log-linear DPO reduce to binary sparse approximation problems that can be solved efficiently by lattice-based and binary matching pursuit methods with recovery guarantees."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem."}],"snapshot_sha256":"0fc296242b5d6b9b57bfa821b2bdaffbc309c1b3095021e6606ad1e709894733"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"3cab39d4d8f1f232287399b979029d5e29bbc975550da0d3ebcd421fd2e15205"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T15:39:43.237812Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-20T03:01:22.782107Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T16:20:23.175153Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.02495/integrity.json","findings":[],"snapshot_sha256":"e123e53f1d28080b036ab00e272a95fc00fabdab7e773726f4130eb401745d9b","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Offline Reinforcement Learning from Human Feedback (RLHF) pipelines such as Direct Preference Optimization (DPO) train on a pre-collected preference dataset, which makes them vulnerable to preference poisoning attack. We study label flip attacks against log-linear DPO. We first illustrate that flipping one preference label induces a parameter-independent shift in the DPO gradient. Using this key property, we can then convert the targeted poisoning problem into a structured binary sparse approximation problem. To solve this problem, we develop two attack methods: Binary-Aware Lattice Attack (BA","authors_text":"Chenye Yang, Lifeng Lai, Weiyu Xu","cross_cats":["cs.AI","stat.ML"],"headline":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T11:45:38Z","title":"Efficient Preference Poisoning Attack on Offline RLHF"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.02495","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T19:25:48.070349Z","id":"4c853f26-0168-484b-a44f-1099298b1503","model_set":{"reader":"grok-4.3"},"one_line_summary":"Label-flip attacks on log-linear DPO reduce to binary sparse approximation problems that can be solved efficiently by lattice-based and binary matching pursuit methods with recovery guarantees.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Flipping one preference label in log-linear DPO creates a parameter-independent gradient shift that turns targeted poisoning into a binary sparse approximation problem.","strongest_claim":"Flipping one preference label induces a parameter-independent shift in the DPO gradient, converting the targeted poisoning problem into a structured binary sparse approximation problem that BAL-A and BMP-A solve with sufficient conditions and coherence-based guarantees.","weakest_assumption":"The central reduction assumes that the DPO objective is strictly log-linear in the parameters so that the gradient shift from a single label flip remains independent of the current parameter vector; if this independence fails for realistic non-linear models or regularizers, the conversion to sparse approximation no longer holds."}},"verdict_id":"4c853f26-0168-484b-a44f-1099298b1503"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:12fa24d1fe832370556d371a27b5623014c63073ddfe9bf4f045f039bfd87fe5","target":"record","created_at":"2026-05-26T02:04:11Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"168897d1319adf9cf5ca6ec100e90635defd2020f0090cbb82ced51119b27a71","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T11:45:38Z","title_canon_sha256":"6ce72836f72c2d4bc21da7421672596e800250bd45781c68f4170bcb702393f2"},"schema_version":"1.0","source":{"id":"2605.02495","kind":"arxiv","version":2}},"canonical_sha256":"983aa4c1b3d94e8851a943226623dcc84bdf1b058985d897d42c7897ef4bb2de","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"983aa4c1b3d94e8851a943226623dcc84bdf1b058985d897d42c7897ef4bb2de","first_computed_at":"2026-05-26T02:04:11.841288Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T02:04:11.841288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ObyHYb7YYsiGUjDOEZPq1Vi0G7mf/VBN5ncMHv0S444cFv7MumSQD9zGaZ9ZA5Aj4+hI8vVsUaCNIvp+jnllBQ==","signature_status":"signed_v1","signed_at":"2026-05-26T02:04:11.842217Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.02495","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:12fa24d1fe832370556d371a27b5623014c63073ddfe9bf4f045f039bfd87fe5","sha256:9f45ddf865fd1f2a21fc66f51b6f6203d3ce0884feb01ddef990eb47159406c9"],"state_sha256":"845d5b804b5875f0fc0b6e596a4e79d67e758a4bfa7096f2ac2d59561b95dee7"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"w39wsaNd54pfCPkO00ypsx7OOhrgI0jD8kaJ464GsOS77tpPurlCgTL5MPJVEOXEM2k25Hq7oiH/9fcMwmR3CQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-31T02:13:50.497464Z","bundle_sha256":"c74c11e222a081680730e0939af7e87ad87b5b3df07ee54920a5117b8dcc6915"}}