{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:65CIL5LUIGETQDWAM2ETXXYYNR","short_pith_number":"pith:65CIL5LU","canonical_record":{"source":{"id":"2502.14643","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T15:30:27Z","cross_cats_sorted":[],"title_canon_sha256":"04e6c59d78ac74c9cf19e7a6e25e54398689d6909a469de60c19f9bffae8ef03","abstract_canon_sha256":"04ed26577e9e66e139363004167e9e1ed2117f3a59303a6a0ca14587c1d99847"},"schema_version":"1.0"},"canonical_sha256":"f74485f5744189380ec066893bdf186c6856130d8b221ab138a3ff9d524a204a","source":{"kind":"arxiv","id":"2502.14643","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.14643","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"arxiv_version","alias_value":"2502.14643v2","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.14643","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_12","alias_value":"65CIL5LUIGET","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_16","alias_value":"65CIL5LUIGETQDWA","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_8","alias_value":"65CIL5LU","created_at":"2026-07-05T11:12:11Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:65CIL5LUIGETQDWAM2ETXXYYNR","target":"record","payload":{"canonical_record":{"source":{"id":"2502.14643","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T15:30:27Z","cross_cats_sorted":[],"title_canon_sha256":"04e6c59d78ac74c9cf19e7a6e25e54398689d6909a469de60c19f9bffae8ef03","abstract_canon_sha256":"04ed26577e9e66e139363004167e9e1ed2117f3a59303a6a0ca14587c1d99847"},"schema_version":"1.0"},"canonical_sha256":"f74485f5744189380ec066893bdf186c6856130d8b221ab138a3ff9d524a204a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:11.142843Z","signature_b64":"GfoYKdVO0RKSc2Ejhw+QSBe03sur8e8C+uPHQg9jYOVktfkNLDZeRRX9bt96FQ/ka7zfRUe9N1vc4fvYUFFWAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f74485f5744189380ec066893bdf186c6856130d8b221ab138a3ff9d524a204a","last_reissued_at":"2026-07-05T11:12:11.142335Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:11.142335Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.14643","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":"2026-07-05T11:12:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1cSjq3mZUS2kY/RVx5CptABAK1If7bCbbsp2wJ0kwdhBuaobWzgz5kKNGR8xbPyawacPexrdfZ98LtmctL2KDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T18:13:06.969358Z"},"content_sha256":"7e2e5c89fb6beca7255e39438b8aaaf7585985dd384ff338756ad356bc2f024f","schema_version":"1.0","event_id":"sha256:7e2e5c89fb6beca7255e39438b8aaaf7585985dd384ff338756ad356bc2f024f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:65CIL5LUIGETQDWAM2ETXXYYNR","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Length-Controlled Margin-Based Preference Optimization without Reference Model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Gengxu Li, Tingyu Xia, Yi Chang, Yuan Wu","submitted_at":"2025-02-20T15:30:27Z","abstract_excerpt":"Direct Preference Optimization (DPO) is a widely adopted offline algorithm for preference-based reinforcement learning from human feedback (RLHF), designed to improve training simplicity and stability by redefining reward functions. However, DPO is hindered by several limitations, including length bias, memory inefficiency, and probability degradation. To address these challenges, we propose Length-Controlled Margin-Based Preference Optimization (LMPO), a more efficient and robust alternative. LMPO introduces a uniform reference model as an upper bound for the DPO loss, enabling a more accurat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.14643","kind":"arxiv","version":2},"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/2502.14643/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":"2026-07-05T11:12:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"rqbjPh+DwVKDWVUopkkrfrQccd36xBofz+NJHrjgCVdl42Cx/4dJqBSKLWIAQ7qAOofBEFfL2zLcOLmhEaywDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T18:13:06.969861Z"},"content_sha256":"f6b0d156bf319b969d773cd761e0f0e7007feaaf5b35e141a761048891d035a7","schema_version":"1.0","event_id":"sha256:f6b0d156bf319b969d773cd761e0f0e7007feaaf5b35e141a761048891d035a7"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/65CIL5LUIGETQDWAM2ETXXYYNR/bundle.json","state_url":"https://pith.science/pith/65CIL5LUIGETQDWAM2ETXXYYNR/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/65CIL5LUIGETQDWAM2ETXXYYNR/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":"2026-08-19T18:13:06Z","links":{"resolver":"https://pith.science/pith/65CIL5LUIGETQDWAM2ETXXYYNR","bundle":"https://pith.science/pith/65CIL5LUIGETQDWAM2ETXXYYNR/bundle.json","state":"https://pith.science/pith/65CIL5LUIGETQDWAM2ETXXYYNR/state.json","well_known_bundle":"https://pith.science/.well-known/pith/65CIL5LUIGETQDWAM2ETXXYYNR/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:65CIL5LUIGETQDWAM2ETXXYYNR","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":"04ed26577e9e66e139363004167e9e1ed2117f3a59303a6a0ca14587c1d99847","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T15:30:27Z","title_canon_sha256":"04e6c59d78ac74c9cf19e7a6e25e54398689d6909a469de60c19f9bffae8ef03"},"schema_version":"1.0","source":{"id":"2502.14643","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.14643","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"arxiv_version","alias_value":"2502.14643v2","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.14643","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_12","alias_value":"65CIL5LUIGET","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_16","alias_value":"65CIL5LUIGETQDWA","created_at":"2026-07-05T11:12:11Z"},{"alias_kind":"pith_short_8","alias_value":"65CIL5LU","created_at":"2026-07-05T11:12:11Z"}],"graph_snapshots":[{"event_id":"sha256:f6b0d156bf319b969d773cd761e0f0e7007feaaf5b35e141a761048891d035a7","target":"graph","created_at":"2026-07-05T11:12:11Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2502.14643/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Direct Preference Optimization (DPO) is a widely adopted offline algorithm for preference-based reinforcement learning from human feedback (RLHF), designed to improve training simplicity and stability by redefining reward functions. However, DPO is hindered by several limitations, including length bias, memory inefficiency, and probability degradation. To address these challenges, we propose Length-Controlled Margin-Based Preference Optimization (LMPO), a more efficient and robust alternative. LMPO introduces a uniform reference model as an upper bound for the DPO loss, enabling a more accurat","authors_text":"Gengxu Li, Tingyu Xia, Yi Chang, Yuan Wu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T15:30:27Z","title":"Length-Controlled Margin-Based Preference Optimization without Reference Model"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.14643","kind":"arxiv","version":2},"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":"sha256:7e2e5c89fb6beca7255e39438b8aaaf7585985dd384ff338756ad356bc2f024f","target":"record","created_at":"2026-07-05T11:12:11Z","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":"04ed26577e9e66e139363004167e9e1ed2117f3a59303a6a0ca14587c1d99847","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T15:30:27Z","title_canon_sha256":"04e6c59d78ac74c9cf19e7a6e25e54398689d6909a469de60c19f9bffae8ef03"},"schema_version":"1.0","source":{"id":"2502.14643","kind":"arxiv","version":2}},"canonical_sha256":"f74485f5744189380ec066893bdf186c6856130d8b221ab138a3ff9d524a204a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f74485f5744189380ec066893bdf186c6856130d8b221ab138a3ff9d524a204a","first_computed_at":"2026-07-05T11:12:11.142335Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:12:11.142335Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"GfoYKdVO0RKSc2Ejhw+QSBe03sur8e8C+uPHQg9jYOVktfkNLDZeRRX9bt96FQ/ka7zfRUe9N1vc4fvYUFFWAg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:12:11.142843Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.14643","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7e2e5c89fb6beca7255e39438b8aaaf7585985dd384ff338756ad356bc2f024f","sha256:f6b0d156bf319b969d773cd761e0f0e7007feaaf5b35e141a761048891d035a7"],"state_sha256":"eef1448351763a6d5e0ce2f3dc9e6ba8ebe8eaaac3110e568a434c312911b4a3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XtmeGi2LL3vlGJ1Jr+W6q60QTvxKPxr4RzPQXSemo/5KO2/tChX/nFgAJudz+T2FX4mutNSltmljD7hHC+qlBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-19T18:13:06.998413Z","bundle_sha256":"283ffdadd0cbfa3e8caa55942218cd62a8c068aa2dfe491bf014b43d6767a679"}}