{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6HFYAO3DBBG6GD3HPHD6F6J5V7","short_pith_number":"pith:6HFYAO3D","schema_version":"1.0","canonical_sha256":"f1cb803b63084de30f6779c7e2f93dafcd8f79d4cb1c1e256cbe2019f1b901e5","source":{"kind":"arxiv","id":"2506.02460","version":1},"attestation_state":"computed","paper":{"title":"MidPO: Dual Preference Optimization for Safety and Helpfulness in Large Language Models via a Mixture of Experts Framework","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Lixin Cui, Lu Bai, Min Yang, Yanlin Wang, Yupeng Qi, Ziyu Lyu","submitted_at":"2025-06-03T05:23:09Z","abstract_excerpt":"As large language models (LLMs) are increasingly applied across various domains, enhancing safety while maintaining the helpfulness of LLMs has become a critical challenge. Recent studies solve this problem through safety-constrained online preference optimization or safety-constrained offline preference optimization. However, the safety-constrained online methods often suffer from excessive safety, which might reduce helpfulness, while the safety-constrained offline methods perform poorly in adaptively balancing safety and helpfulness. To address these limitations, we propose MidPO, a \\textbf"},"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":"2506.02460","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-03T05:23:09Z","cross_cats_sorted":[],"title_canon_sha256":"7ee10ed983c660502566027e8fef6e22f2411e18d113a42c68cd89a3dc7464e9","abstract_canon_sha256":"2b0fbcf865b52ab9bfc8605bce205434a0d2a6a9d6bec9ec191ccdc4c6727217"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:15:02.374081Z","signature_b64":"i71ZdWARtHVnbosF3JHnXnJ0H5361l/F9GrE1kAFUwREa4x3sxOjiratDsSVANJLtgD9LP34pysUmeomRrdSDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f1cb803b63084de30f6779c7e2f93dafcd8f79d4cb1c1e256cbe2019f1b901e5","last_reissued_at":"2026-07-05T11:15:02.373579Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:15:02.373579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MidPO: Dual Preference Optimization for Safety and Helpfulness in Large Language Models via a Mixture of Experts Framework","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Lixin Cui, Lu Bai, Min Yang, Yanlin Wang, Yupeng Qi, Ziyu Lyu","submitted_at":"2025-06-03T05:23:09Z","abstract_excerpt":"As large language models (LLMs) are increasingly applied across various domains, enhancing safety while maintaining the helpfulness of LLMs has become a critical challenge. Recent studies solve this problem through safety-constrained online preference optimization or safety-constrained offline preference optimization. However, the safety-constrained online methods often suffer from excessive safety, which might reduce helpfulness, while the safety-constrained offline methods perform poorly in adaptively balancing safety and helpfulness. To address these limitations, we propose MidPO, a \\textbf"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.02460","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/2506.02460/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":"2506.02460","created_at":"2026-07-05T11:15:02.373645+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.02460v1","created_at":"2026-07-05T11:15:02.373645+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.02460","created_at":"2026-07-05T11:15:02.373645+00:00"},{"alias_kind":"pith_short_12","alias_value":"6HFYAO3DBBG6","created_at":"2026-07-05T11:15:02.373645+00:00"},{"alias_kind":"pith_short_16","alias_value":"6HFYAO3DBBG6GD3H","created_at":"2026-07-05T11:15:02.373645+00:00"},{"alias_kind":"pith_short_8","alias_value":"6HFYAO3D","created_at":"2026-07-05T11:15:02.373645+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7","json":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7.json","graph_json":"https://pith.science/api/pith-number/6HFYAO3DBBG6GD3HPHD6F6J5V7/graph.json","events_json":"https://pith.science/api/pith-number/6HFYAO3DBBG6GD3HPHD6F6J5V7/events.json","paper":"https://pith.science/paper/6HFYAO3D"},"agent_actions":{"view_html":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7","download_json":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7.json","view_paper":"https://pith.science/paper/6HFYAO3D","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.02460&json=true","fetch_graph":"https://pith.science/api/pith-number/6HFYAO3DBBG6GD3HPHD6F6J5V7/graph.json","fetch_events":"https://pith.science/api/pith-number/6HFYAO3DBBG6GD3HPHD6F6J5V7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7/action/storage_attestation","attest_author":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7/action/author_attestation","sign_citation":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7/action/citation_signature","submit_replication":"https://pith.science/pith/6HFYAO3DBBG6GD3HPHD6F6J5V7/action/replication_record"}},"created_at":"2026-07-05T11:15:02.373645+00:00","updated_at":"2026-07-05T11:15:02.373645+00:00"}