{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:CEUDEHRPCMFKLJ4DYQUCEQO74Z","short_pith_number":"pith:CEUDEHRP","canonical_record":{"source":{"id":"2506.11098","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T13:19:07Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"64bb4c76adac624ca0313c99270c8d62d70cbba7ea4559ae629b61baea77a373","abstract_canon_sha256":"2ac16ff4c22232a58ada5882942fb742a758de533031f625edcacff69a78e985"},"schema_version":"1.0"},"canonical_sha256":"1128321e2f130aa5a783c4282241dfe679a9d10d8c2472c79e66ce6bf97d59b5","source":{"kind":"arxiv","id":"2506.11098","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.11098","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"arxiv_version","alias_value":"2506.11098v1","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.11098","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_12","alias_value":"CEUDEHRPCMFK","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_16","alias_value":"CEUDEHRPCMFKLJ4D","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_8","alias_value":"CEUDEHRP","created_at":"2026-07-05T11:20:59Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:CEUDEHRPCMFKLJ4DYQUCEQO74Z","target":"record","payload":{"canonical_record":{"source":{"id":"2506.11098","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T13:19:07Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"64bb4c76adac624ca0313c99270c8d62d70cbba7ea4559ae629b61baea77a373","abstract_canon_sha256":"2ac16ff4c22232a58ada5882942fb742a758de533031f625edcacff69a78e985"},"schema_version":"1.0"},"canonical_sha256":"1128321e2f130aa5a783c4282241dfe679a9d10d8c2472c79e66ce6bf97d59b5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:59.350177Z","signature_b64":"o3U5Yqt7e+WlYiJu+nWJPJ/The7fbe6XlnUXl0zJ8UiNZQqrNLbZtxGxxZlEiKj36uH8Eq6Y7h4W+p6+xrnvCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1128321e2f130aa5a783c4282241dfe679a9d10d8c2472c79e66ce6bf97d59b5","last_reissued_at":"2026-07-05T11:20:59.349803Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:59.349803Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2506.11098","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:20:59Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PE+Vuz/fT44IxNpvZtkdv9qsrG6Jd0SL6PEg8Md/8PCPyLnDUamKF8DF1CobaYmSFyAHWpMGHRPbbQsXvAMRDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T04:10:19.663843Z"},"content_sha256":"0710b2d61fc614a710e36afa5531d88ac2fab8510234e0f6609ee15504b23601","schema_version":"1.0","event_id":"sha256:0710b2d61fc614a710e36afa5531d88ac2fab8510234e0f6609ee15504b23601"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:CEUDEHRPCMFKLJ4DYQUCEQO74Z","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Debiasing Online Preference Learning via Preference Feature Preservation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Dongyoung Kim, Jaehyung Kim, Jinsung Yoon, Jinwoo Shin","submitted_at":"2025-06-06T13:19:07Z","abstract_excerpt":"Recent preference learning frameworks for large language models (LLMs) simplify human preferences with binary pairwise comparisons and scalar rewards. This simplification could make LLMs' responses biased to mostly preferred features, and would be exacerbated during the iterations of online preference learning steps. To address these challenges, we propose a novel framework coined PFP (Preference Feature Preservation). The key idea of PFP is maintaining the distribution of human preference features and utilizing such rich signals throughout the online preference learning process. Specifically,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.11098","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/2506.11098/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:20:59Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YXnx8ycNH7ENST0YD3Gq6BpcpBVs+LnJuZwr84Lo4cbj1NH+dsPHXu3bM2aOt4rHYoIeBZheyaq/5Ah9BfiFBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T04:10:19.664354Z"},"content_sha256":"d68773d4f1c0c48d89afc4dde116966c0def2b8a28435ca0ccba2374a4959c36","schema_version":"1.0","event_id":"sha256:d68773d4f1c0c48d89afc4dde116966c0def2b8a28435ca0ccba2374a4959c36"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/bundle.json","state_url":"https://pith.science/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/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":"2026-08-08T04:10:19Z","links":{"resolver":"https://pith.science/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z","bundle":"https://pith.science/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/bundle.json","state":"https://pith.science/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/state.json","well_known_bundle":"https://pith.science/.well-known/pith/CEUDEHRPCMFKLJ4DYQUCEQO74Z/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:CEUDEHRPCMFKLJ4DYQUCEQO74Z","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":"2ac16ff4c22232a58ada5882942fb742a758de533031f625edcacff69a78e985","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T13:19:07Z","title_canon_sha256":"64bb4c76adac624ca0313c99270c8d62d70cbba7ea4559ae629b61baea77a373"},"schema_version":"1.0","source":{"id":"2506.11098","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.11098","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"arxiv_version","alias_value":"2506.11098v1","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.11098","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_12","alias_value":"CEUDEHRPCMFK","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_16","alias_value":"CEUDEHRPCMFKLJ4D","created_at":"2026-07-05T11:20:59Z"},{"alias_kind":"pith_short_8","alias_value":"CEUDEHRP","created_at":"2026-07-05T11:20:59Z"}],"graph_snapshots":[{"event_id":"sha256:d68773d4f1c0c48d89afc4dde116966c0def2b8a28435ca0ccba2374a4959c36","target":"graph","created_at":"2026-07-05T11:20:59Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2506.11098/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent preference learning frameworks for large language models (LLMs) simplify human preferences with binary pairwise comparisons and scalar rewards. This simplification could make LLMs' responses biased to mostly preferred features, and would be exacerbated during the iterations of online preference learning steps. To address these challenges, we propose a novel framework coined PFP (Preference Feature Preservation). The key idea of PFP is maintaining the distribution of human preference features and utilizing such rich signals throughout the online preference learning process. Specifically,","authors_text":"Dongyoung Kim, Jaehyung Kim, Jinsung Yoon, Jinwoo Shin","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T13:19:07Z","title":"Debiasing Online Preference Learning via Preference Feature Preservation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.11098","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:0710b2d61fc614a710e36afa5531d88ac2fab8510234e0f6609ee15504b23601","target":"record","created_at":"2026-07-05T11:20:59Z","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":"2ac16ff4c22232a58ada5882942fb742a758de533031f625edcacff69a78e985","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T13:19:07Z","title_canon_sha256":"64bb4c76adac624ca0313c99270c8d62d70cbba7ea4559ae629b61baea77a373"},"schema_version":"1.0","source":{"id":"2506.11098","kind":"arxiv","version":1}},"canonical_sha256":"1128321e2f130aa5a783c4282241dfe679a9d10d8c2472c79e66ce6bf97d59b5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"1128321e2f130aa5a783c4282241dfe679a9d10d8c2472c79e66ce6bf97d59b5","first_computed_at":"2026-07-05T11:20:59.349803Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:20:59.349803Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"o3U5Yqt7e+WlYiJu+nWJPJ/The7fbe6XlnUXl0zJ8UiNZQqrNLbZtxGxxZlEiKj36uH8Eq6Y7h4W+p6+xrnvCg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:20:59.350177Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.11098","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:0710b2d61fc614a710e36afa5531d88ac2fab8510234e0f6609ee15504b23601","sha256:d68773d4f1c0c48d89afc4dde116966c0def2b8a28435ca0ccba2374a4959c36"],"state_sha256":"7e75c5603df7732383d8ca341cc77fdc8c4db7e997e431c258d02ac2ae7c5f6b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mpGZ4bN1GbjlhNRGqRjZ6e/SbdWHZq24nJxpdg5zsNxFlmSYbAqb51W8kyBRDQvtGXTATQqVlnSunQ6gJa7xDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T04:10:19.667753Z","bundle_sha256":"57732685dd1bb824cacece775ccbe180795445cdf6299ef248e3b6308f642816"}}