{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JF3BXD3Y2QYCZ6AHCS57UQCBMN","short_pith_number":"pith:JF3BXD3Y","schema_version":"1.0","canonical_sha256":"49761b8f78d4302cf80714bbfa404163673323e31be23a2fbf7353964c17a4e1","source":{"kind":"arxiv","id":"2602.02495","version":3},"attestation_state":"computed","paper":{"title":"Reward-free Alignment for Conflicting Objectives","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Peter Chen, Tianyi Lin, Xiaopeng Li, Xi Chen","submitted_at":"2026-02-02T18:59:52Z","abstract_excerpt":"Direct alignment methods are increasingly used to align large language models (LLMs) with human preferences. However, many real-world alignment problems involve multiple conflicting objectives, where naive aggregation of preferences can lead to unstable training and poor trade-offs. In particular, weighted loss methods may fail to identify update directions that simultaneously improve all objectives, and existing multi-objective approaches often rely on explicit reward models, introducing additional complexity and distorting user-specified preferences. The contributions of this paper are two-f"},"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":"2602.02495","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-02-02T18:59:52Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"e8db16f24dc3c1737b6b87c320a7f324d47c466c5ff8565b64ca537adf01b224","abstract_canon_sha256":"cef4cc4d0159a9652e9d498a0a92795022024f8e079d09c704388675c88c5238"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:04.451265Z","signature_b64":"8jcR8SyL9kiUHrJbZx1tipWJk2OVxFr/ZhLrZko+IIKSwIYHqQhochYWR2Uom6+yH/7C/Bmsgpz46MDyHQSvCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"49761b8f78d4302cf80714bbfa404163673323e31be23a2fbf7353964c17a4e1","last_reissued_at":"2026-05-26T02:04:04.450336Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:04.450336Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reward-free Alignment for Conflicting Objectives","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Peter Chen, Tianyi Lin, Xiaopeng Li, Xi Chen","submitted_at":"2026-02-02T18:59:52Z","abstract_excerpt":"Direct alignment methods are increasingly used to align large language models (LLMs) with human preferences. However, many real-world alignment problems involve multiple conflicting objectives, where naive aggregation of preferences can lead to unstable training and poor trade-offs. In particular, weighted loss methods may fail to identify update directions that simultaneously improve all objectives, and existing multi-objective approaches often rely on explicit reward models, introducing additional complexity and distorting user-specified preferences. The contributions of this paper are two-f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.02495","kind":"arxiv","version":3},"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/2602.02495/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":"2602.02495","created_at":"2026-05-26T02:04:04.450486+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.02495v3","created_at":"2026-05-26T02:04:04.450486+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.02495","created_at":"2026-05-26T02:04:04.450486+00:00"},{"alias_kind":"pith_short_12","alias_value":"JF3BXD3Y2QYC","created_at":"2026-05-26T02:04:04.450486+00:00"},{"alias_kind":"pith_short_16","alias_value":"JF3BXD3Y2QYCZ6AH","created_at":"2026-05-26T02:04:04.450486+00:00"},{"alias_kind":"pith_short_8","alias_value":"JF3BXD3Y","created_at":"2026-05-26T02:04:04.450486+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.09975","citing_title":"Chebyshev Center-Based Direction Selection for Multi-Objective Optimization and Training PINNs","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05750","citing_title":"RVPO: Risk-Sensitive Alignment via Variance Regularization","ref_index":24,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN","json":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN.json","graph_json":"https://pith.science/api/pith-number/JF3BXD3Y2QYCZ6AHCS57UQCBMN/graph.json","events_json":"https://pith.science/api/pith-number/JF3BXD3Y2QYCZ6AHCS57UQCBMN/events.json","paper":"https://pith.science/paper/JF3BXD3Y"},"agent_actions":{"view_html":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN","download_json":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN.json","view_paper":"https://pith.science/paper/JF3BXD3Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.02495&json=true","fetch_graph":"https://pith.science/api/pith-number/JF3BXD3Y2QYCZ6AHCS57UQCBMN/graph.json","fetch_events":"https://pith.science/api/pith-number/JF3BXD3Y2QYCZ6AHCS57UQCBMN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN/action/storage_attestation","attest_author":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN/action/author_attestation","sign_citation":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN/action/citation_signature","submit_replication":"https://pith.science/pith/JF3BXD3Y2QYCZ6AHCS57UQCBMN/action/replication_record"}},"created_at":"2026-05-26T02:04:04.450486+00:00","updated_at":"2026-05-26T02:04:04.450486+00:00"}