{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:42DMKMMTVC5IDHGSVRIGVUU2OV","short_pith_number":"pith:42DMKMMT","schema_version":"1.0","canonical_sha256":"e686c53193a8ba819cd2ac506ad29a754cd71bb60d3093142235a6e57a6bd124","source":{"kind":"arxiv","id":"2508.07750","version":1},"attestation_state":"computed","paper":{"title":"Learning to Align, Aligning to Learn: A Unified Approach for Self-Optimized Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Cheng Wei, Chunxiao Guo, Haowen Wang, Jiadi Jiang, Jian Wang, Jiaxin Liang, Ji Li, Jingyuan Deng, Jinjie Gu, Lei Fan, Peng Wei, Shuowen Zhang, Xudong Han, Yun Yue, Zhiling Ye","submitted_at":"2025-08-11T08:28:47Z","abstract_excerpt":"Alignment methodologies have emerged as a critical pathway for enhancing language model alignment capabilities. While SFT (supervised fine-tuning) accelerates convergence through direct token-level loss intervention, its efficacy is constrained by offline policy trajectory. In contrast, RL(reinforcement learning) facilitates exploratory policy optimization, but suffers from low sample efficiency and stringent dependency on high-quality base models. To address these dual challenges, we propose GRAO (Group Relative Alignment Optimization), a unified framework that synergizes the respective stren"},"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":"2508.07750","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-08-11T08:28:47Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"745f526e6d20977595262bba3861899911ab12e9b873b32d14a953621f7f99fc","abstract_canon_sha256":"1c42b27c243a6336c2aa8df9ea907d0e1465ad555f898f379b39a81dd641f7f0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:51:51.603251Z","signature_b64":"NboXR/0qtpWEnPhAt1iNt78o4veYqhRe1p4S7MCpFOYzTRA1xrrT4GInmifzEhh+nlRoQHOUMkT9Wn6KplexAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e686c53193a8ba819cd2ac506ad29a754cd71bb60d3093142235a6e57a6bd124","last_reissued_at":"2026-07-05T11:51:51.602798Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:51:51.602798Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to Align, Aligning to Learn: A Unified Approach for Self-Optimized Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Cheng Wei, Chunxiao Guo, Haowen Wang, Jiadi Jiang, Jian Wang, Jiaxin Liang, Ji Li, Jingyuan Deng, Jinjie Gu, Lei Fan, Peng Wei, Shuowen Zhang, Xudong Han, Yun Yue, Zhiling Ye","submitted_at":"2025-08-11T08:28:47Z","abstract_excerpt":"Alignment methodologies have emerged as a critical pathway for enhancing language model alignment capabilities. While SFT (supervised fine-tuning) accelerates convergence through direct token-level loss intervention, its efficacy is constrained by offline policy trajectory. In contrast, RL(reinforcement learning) facilitates exploratory policy optimization, but suffers from low sample efficiency and stringent dependency on high-quality base models. To address these dual challenges, we propose GRAO (Group Relative Alignment Optimization), a unified framework that synergizes the respective stren"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.07750","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/2508.07750/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":"2508.07750","created_at":"2026-07-05T11:51:51.602856+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.07750v1","created_at":"2026-07-05T11:51:51.602856+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.07750","created_at":"2026-07-05T11:51:51.602856+00:00"},{"alias_kind":"pith_short_12","alias_value":"42DMKMMTVC5I","created_at":"2026-07-05T11:51:51.602856+00:00"},{"alias_kind":"pith_short_16","alias_value":"42DMKMMTVC5IDHGS","created_at":"2026-07-05T11:51:51.602856+00:00"},{"alias_kind":"pith_short_8","alias_value":"42DMKMMT","created_at":"2026-07-05T11:51:51.602856+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.07941","citing_title":"Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning","ref_index":120,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV","json":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV.json","graph_json":"https://pith.science/api/pith-number/42DMKMMTVC5IDHGSVRIGVUU2OV/graph.json","events_json":"https://pith.science/api/pith-number/42DMKMMTVC5IDHGSVRIGVUU2OV/events.json","paper":"https://pith.science/paper/42DMKMMT"},"agent_actions":{"view_html":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV","download_json":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV.json","view_paper":"https://pith.science/paper/42DMKMMT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.07750&json=true","fetch_graph":"https://pith.science/api/pith-number/42DMKMMTVC5IDHGSVRIGVUU2OV/graph.json","fetch_events":"https://pith.science/api/pith-number/42DMKMMTVC5IDHGSVRIGVUU2OV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV/action/storage_attestation","attest_author":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV/action/author_attestation","sign_citation":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV/action/citation_signature","submit_replication":"https://pith.science/pith/42DMKMMTVC5IDHGSVRIGVUU2OV/action/replication_record"}},"created_at":"2026-07-05T11:51:51.602856+00:00","updated_at":"2026-07-05T11:51:51.602856+00:00"}