{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:M7PG2TQVHYVXNARPW6CZSFBRGA","short_pith_number":"pith:M7PG2TQV","schema_version":"1.0","canonical_sha256":"67de6d4e153e2b76822fb78599143130388051703433f62200e51ddadb2dd7f6","source":{"kind":"arxiv","id":"2406.15178","version":1},"attestation_state":"computed","paper":{"title":"Hybrid Alignment Training for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bei Li, Chenglong Wang, Hang Zhou, Jingbo Zhu, Kaiyan Chang, Tongran Liu, Tong Xiao, Yongyu Mu","submitted_at":"2024-06-21T14:23:57Z","abstract_excerpt":"Alignment training is crucial for enabling large language models (LLMs) to cater to human intentions and preferences. It is typically performed based on two stages with different objectives: instruction-following alignment and human-preference alignment. However, aligning LLMs with these objectives in sequence suffers from an inherent problem: the objectives may conflict, and the LLMs cannot guarantee to simultaneously align with the instructions and human preferences well. To response to these, in this work, we propose a Hybrid Alignment Training (Hbat) approach, based on alternating alignmen"},"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":"2406.15178","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-21T14:23:57Z","cross_cats_sorted":[],"title_canon_sha256":"d506afd100340f3ba131e5a4b935bcb7b62e76c4c57c5ce46f6f35c9e42c629c","abstract_canon_sha256":"da2c792b1a4f6433da75ee56efa093c7a4ed4440b6ea01fd43c4fe94a5450f8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:35:15.046644Z","signature_b64":"xeLvSLQZxb2GWmogpPBrqJO46oD2WU74OAdVpyuZ4G4D078nPalqZq6enzhCcxNR0vynC+kdvYRJs1YxunbADg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"67de6d4e153e2b76822fb78599143130388051703433f62200e51ddadb2dd7f6","last_reissued_at":"2026-07-05T08:35:15.046184Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:35:15.046184Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hybrid Alignment Training for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bei Li, Chenglong Wang, Hang Zhou, Jingbo Zhu, Kaiyan Chang, Tongran Liu, Tong Xiao, Yongyu Mu","submitted_at":"2024-06-21T14:23:57Z","abstract_excerpt":"Alignment training is crucial for enabling large language models (LLMs) to cater to human intentions and preferences. It is typically performed based on two stages with different objectives: instruction-following alignment and human-preference alignment. However, aligning LLMs with these objectives in sequence suffers from an inherent problem: the objectives may conflict, and the LLMs cannot guarantee to simultaneously align with the instructions and human preferences well. To response to these, in this work, we propose a Hybrid Alignment Training (Hbat) approach, based on alternating alignmen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.15178","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/2406.15178/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":"2406.15178","created_at":"2026-07-05T08:35:15.046253+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.15178v1","created_at":"2026-07-05T08:35:15.046253+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.15178","created_at":"2026-07-05T08:35:15.046253+00:00"},{"alias_kind":"pith_short_12","alias_value":"M7PG2TQVHYVX","created_at":"2026-07-05T08:35:15.046253+00:00"},{"alias_kind":"pith_short_16","alias_value":"M7PG2TQVHYVXNARP","created_at":"2026-07-05T08:35:15.046253+00:00"},{"alias_kind":"pith_short_8","alias_value":"M7PG2TQV","created_at":"2026-07-05T08:35:15.046253+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.19922","citing_title":"HEAL: A Hypothesis-Based Preference-Aware Analysis Framework","ref_index":25,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA","json":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA.json","graph_json":"https://pith.science/api/pith-number/M7PG2TQVHYVXNARPW6CZSFBRGA/graph.json","events_json":"https://pith.science/api/pith-number/M7PG2TQVHYVXNARPW6CZSFBRGA/events.json","paper":"https://pith.science/paper/M7PG2TQV"},"agent_actions":{"view_html":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA","download_json":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA.json","view_paper":"https://pith.science/paper/M7PG2TQV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.15178&json=true","fetch_graph":"https://pith.science/api/pith-number/M7PG2TQVHYVXNARPW6CZSFBRGA/graph.json","fetch_events":"https://pith.science/api/pith-number/M7PG2TQVHYVXNARPW6CZSFBRGA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA/action/storage_attestation","attest_author":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA/action/author_attestation","sign_citation":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA/action/citation_signature","submit_replication":"https://pith.science/pith/M7PG2TQVHYVXNARPW6CZSFBRGA/action/replication_record"}},"created_at":"2026-07-05T08:35:15.046253+00:00","updated_at":"2026-07-05T08:35:15.046253+00:00"}