{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:L6L3YHU5QIQUM5OL4GJ4T6OMAV","short_pith_number":"pith:L6L3YHU5","schema_version":"1.0","canonical_sha256":"5f97bc1e9d82214675cbe193c9f9cc05598e1ee89d429563671cb27c5cc7329f","source":{"kind":"arxiv","id":"2406.08124","version":2},"attestation_state":"computed","paper":{"title":"Legend: Leveraging Representation Engineering to Annotate Safety Margin for Preference Datasets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bowen Qin, Chen Huang, Duanyu Feng, Wenqiang Lei, Youcheng Huang, Zheng Zhang","submitted_at":"2024-06-12T12:06:32Z","abstract_excerpt":"The success of the reward model in distinguishing between responses with subtle safety differences depends critically on the high-quality preference dataset, which should capture the fine-grained nuances of harmful and harmless responses. This motivates the need to develop a dataset involving preference margins, which accurately quantify how harmless one response is compared to another. In this paper, we take the first step to propose an effective and cost-efficient framework to promote the margin-enhanced preference dataset development. Our framework, Legend, Leverages representation engineer"},"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":"2406.08124","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-12T12:06:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"23c2fcdde08e5fc96adf243d9acc9c079aa8a0f555cde2932d6bb80a3c5fa824","abstract_canon_sha256":"de33d0c138cb4b19743ca124ceb86e7a39ab397cd769e7757920ef179c7fd473"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:50:48.414287Z","signature_b64":"SjuBOKW+JRCkHZatk3CdGM0JS7ngsDa5x5lpqBdUpBDPuMfn1+EnW/zIEB1WxLltJDJOOsBociEzTe5aStJmBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5f97bc1e9d82214675cbe193c9f9cc05598e1ee89d429563671cb27c5cc7329f","last_reissued_at":"2026-07-05T09:50:48.413736Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:50:48.413736Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Legend: Leveraging Representation Engineering to Annotate Safety Margin for Preference Datasets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bowen Qin, Chen Huang, Duanyu Feng, Wenqiang Lei, Youcheng Huang, Zheng Zhang","submitted_at":"2024-06-12T12:06:32Z","abstract_excerpt":"The success of the reward model in distinguishing between responses with subtle safety differences depends critically on the high-quality preference dataset, which should capture the fine-grained nuances of harmful and harmless responses. This motivates the need to develop a dataset involving preference margins, which accurately quantify how harmless one response is compared to another. In this paper, we take the first step to propose an effective and cost-efficient framework to promote the margin-enhanced preference dataset development. Our framework, Legend, Leverages representation engineer"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.08124","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/2406.08124/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":"2406.08124","created_at":"2026-07-05T09:50:48.413799+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.08124v2","created_at":"2026-07-05T09:50:48.413799+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.08124","created_at":"2026-07-05T09:50:48.413799+00:00"},{"alias_kind":"pith_short_12","alias_value":"L6L3YHU5QIQU","created_at":"2026-07-05T09:50:48.413799+00:00"},{"alias_kind":"pith_short_16","alias_value":"L6L3YHU5QIQUM5OL","created_at":"2026-07-05T09:50:48.413799+00:00"},{"alias_kind":"pith_short_8","alias_value":"L6L3YHU5","created_at":"2026-07-05T09:50:48.413799+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2511.10946","citing_title":"Abstract 3D Perception for Spatial Intelligence in Vision-Language Models","ref_index":39,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV","json":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV.json","graph_json":"https://pith.science/api/pith-number/L6L3YHU5QIQUM5OL4GJ4T6OMAV/graph.json","events_json":"https://pith.science/api/pith-number/L6L3YHU5QIQUM5OL4GJ4T6OMAV/events.json","paper":"https://pith.science/paper/L6L3YHU5"},"agent_actions":{"view_html":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV","download_json":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV.json","view_paper":"https://pith.science/paper/L6L3YHU5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.08124&json=true","fetch_graph":"https://pith.science/api/pith-number/L6L3YHU5QIQUM5OL4GJ4T6OMAV/graph.json","fetch_events":"https://pith.science/api/pith-number/L6L3YHU5QIQUM5OL4GJ4T6OMAV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV/action/storage_attestation","attest_author":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV/action/author_attestation","sign_citation":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV/action/citation_signature","submit_replication":"https://pith.science/pith/L6L3YHU5QIQUM5OL4GJ4T6OMAV/action/replication_record"}},"created_at":"2026-07-05T09:50:48.413799+00:00","updated_at":"2026-07-05T09:50:48.413799+00:00"}