{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:GQ4MQ2BACHOIPEC2HVCKPCQ5BI","short_pith_number":"pith:GQ4MQ2BA","canonical_record":{"source":{"id":"2407.09072","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-12T07:52:32Z","cross_cats_sorted":[],"title_canon_sha256":"c1ea585f4e8a2121f56fe4e2c10aa5df3348c528a0473c300d421b2cc7000af4","abstract_canon_sha256":"116fad6b05b402b975636002b6fcb95a252b9dc8f44a95ee175105d17198677d"},"schema_version":"1.0"},"canonical_sha256":"3438c8682011dc87905a3d44a78a1d0a1ea338a8ae4c62c2f9daa93d99198738","source":{"kind":"arxiv","id":"2407.09072","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.09072","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"arxiv_version","alias_value":"2407.09072v1","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.09072","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_12","alias_value":"GQ4MQ2BACHOI","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_16","alias_value":"GQ4MQ2BACHOIPEC2","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_8","alias_value":"GQ4MQ2BA","created_at":"2026-07-05T08:43:14Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:GQ4MQ2BACHOIPEC2HVCKPCQ5BI","target":"record","payload":{"canonical_record":{"source":{"id":"2407.09072","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-12T07:52:32Z","cross_cats_sorted":[],"title_canon_sha256":"c1ea585f4e8a2121f56fe4e2c10aa5df3348c528a0473c300d421b2cc7000af4","abstract_canon_sha256":"116fad6b05b402b975636002b6fcb95a252b9dc8f44a95ee175105d17198677d"},"schema_version":"1.0"},"canonical_sha256":"3438c8682011dc87905a3d44a78a1d0a1ea338a8ae4c62c2f9daa93d99198738","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:43:14.224396Z","signature_b64":"oBXmBhCHdkVrOg1gT5DXSDVPrP13R4ZvC+nDM8b2HOHpPSKHOghGDQeAWVfmSamd3ETzS2pYw11XPt5OxUJiBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3438c8682011dc87905a3d44a78a1d0a1ea338a8ae4c62c2f9daa93d99198738","last_reissued_at":"2026-07-05T08:43:14.223890Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:43:14.223890Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2407.09072","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":"2026-07-05T08:43:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"D5B9q2s//Sd0G2J1GvdBvHLOBWO74C6/jTEIXFkX78m+OzrXH7nSTz88/Pg6P10u0k/Vf8fFczI3XI45kOy3CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T18:01:45.539020Z"},"content_sha256":"31e79fb93e53199cb1272f9f7b9638e47bf531f451824e8ef3b3e5216b339378","schema_version":"1.0","event_id":"sha256:31e79fb93e53199cb1272f9f7b9638e47bf531f451824e8ef3b3e5216b339378"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:GQ4MQ2BACHOIPEC2HVCKPCQ5BI","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"New Desiderata for Direct Preference Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"David Wipf, Tong He, Xiangkun Hu","submitted_at":"2024-07-12T07:52:32Z","abstract_excerpt":"Large language models in the past have typically relied on some form of reinforcement learning with human feedback (RLHF) to better align model responses with human preferences. However, because of oft-observed instabilities when implementing these RLHF pipelines, various reparameterization techniques have recently been introduced to sidestep the need for separately learning an RL reward model. Instead, directly fine-tuning for human preferences is achieved via the minimization of a single closed-form training objective, a process originally referred to as direct preference optimization (DPO) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.09072","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/2407.09072/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":"2026-07-05T08:43:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"42VqZsWtftuKLujVSZD/P0yoQlqBblDAIF9/+omtTpwSRKNyNMCRRIGa3IESt2TZXmjV5WACNyGZjomt2P+5Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T18:01:45.539563Z"},"content_sha256":"f3fc833d4d8fded46cc37dc2c62dbdccd2db83af7e630505b4b02c0ecc37b194","schema_version":"1.0","event_id":"sha256:f3fc833d4d8fded46cc37dc2c62dbdccd2db83af7e630505b4b02c0ecc37b194"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/bundle.json","state_url":"https://pith.science/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/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":"2026-08-11T18:01:45Z","links":{"resolver":"https://pith.science/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI","bundle":"https://pith.science/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/bundle.json","state":"https://pith.science/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GQ4MQ2BACHOIPEC2HVCKPCQ5BI/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:GQ4MQ2BACHOIPEC2HVCKPCQ5BI","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":"116fad6b05b402b975636002b6fcb95a252b9dc8f44a95ee175105d17198677d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-12T07:52:32Z","title_canon_sha256":"c1ea585f4e8a2121f56fe4e2c10aa5df3348c528a0473c300d421b2cc7000af4"},"schema_version":"1.0","source":{"id":"2407.09072","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.09072","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"arxiv_version","alias_value":"2407.09072v1","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.09072","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_12","alias_value":"GQ4MQ2BACHOI","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_16","alias_value":"GQ4MQ2BACHOIPEC2","created_at":"2026-07-05T08:43:14Z"},{"alias_kind":"pith_short_8","alias_value":"GQ4MQ2BA","created_at":"2026-07-05T08:43:14Z"}],"graph_snapshots":[{"event_id":"sha256:f3fc833d4d8fded46cc37dc2c62dbdccd2db83af7e630505b4b02c0ecc37b194","target":"graph","created_at":"2026-07-05T08:43:14Z","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/2407.09072/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models in the past have typically relied on some form of reinforcement learning with human feedback (RLHF) to better align model responses with human preferences. However, because of oft-observed instabilities when implementing these RLHF pipelines, various reparameterization techniques have recently been introduced to sidestep the need for separately learning an RL reward model. Instead, directly fine-tuning for human preferences is achieved via the minimization of a single closed-form training objective, a process originally referred to as direct preference optimization (DPO) ","authors_text":"David Wipf, Tong He, Xiangkun Hu","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-12T07:52:32Z","title":"New Desiderata for Direct Preference Optimization"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.09072","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":"sha256:31e79fb93e53199cb1272f9f7b9638e47bf531f451824e8ef3b3e5216b339378","target":"record","created_at":"2026-07-05T08:43:14Z","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":"116fad6b05b402b975636002b6fcb95a252b9dc8f44a95ee175105d17198677d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-12T07:52:32Z","title_canon_sha256":"c1ea585f4e8a2121f56fe4e2c10aa5df3348c528a0473c300d421b2cc7000af4"},"schema_version":"1.0","source":{"id":"2407.09072","kind":"arxiv","version":1}},"canonical_sha256":"3438c8682011dc87905a3d44a78a1d0a1ea338a8ae4c62c2f9daa93d99198738","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3438c8682011dc87905a3d44a78a1d0a1ea338a8ae4c62c2f9daa93d99198738","first_computed_at":"2026-07-05T08:43:14.223890Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:43:14.223890Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"oBXmBhCHdkVrOg1gT5DXSDVPrP13R4ZvC+nDM8b2HOHpPSKHOghGDQeAWVfmSamd3ETzS2pYw11XPt5OxUJiBQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:43:14.224396Z","signed_message":"canonical_sha256_bytes"},"source_id":"2407.09072","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:31e79fb93e53199cb1272f9f7b9638e47bf531f451824e8ef3b3e5216b339378","sha256:f3fc833d4d8fded46cc37dc2c62dbdccd2db83af7e630505b4b02c0ecc37b194"],"state_sha256":"22c25d356a05ea8c1044f774340d6c224d9cd5bb2136c63dec44d09ef53bf85c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"m5RlF1zdKZwy16KFwXqD7IK7T/hgVNxw0fchnKJ/ftwv5tfinYXFYECmPZrtBpcxwhQfswWJmkL/JMNTL0tPDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-11T18:01:45.545709Z","bundle_sha256":"564c0479de2aa79f16b1ede1dbcefcd6786c73cf6a2e62775730350d785034eb"}}