{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ALKGED3GELRH62R4BWYPR5M2WZ","short_pith_number":"pith:ALKGED3G","schema_version":"1.0","canonical_sha256":"02d4620f6622e27f6a3c0db0f8f59ab6750aceb75fd8de4e174434e04cada5d6","source":{"kind":"arxiv","id":"2404.04291","version":1},"attestation_state":"computed","paper":{"title":"Investigating Regularization of Self-Play Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Abdalgader Abubaker, Mastane Achab, Mohamed El Amine Seddik, Reda Alami, Salem Lahlou","submitted_at":"2024-04-04T05:38:44Z","abstract_excerpt":"This paper explores the effects of various forms of regularization in the context of language model alignment via self-play. While both reinforcement learning from human feedback (RLHF) and direct preference optimization (DPO) require to collect costly human-annotated pairwise preferences, the self-play fine-tuning (SPIN) approach replaces the rejected answers by data generated from the previous iterate. However, the SPIN method presents a performance instability issue in the learning phase, which can be mitigated by playing against a mixture of the two previous iterates. In the same vein, we "},"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":"2404.04291","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-04-04T05:38:44Z","cross_cats_sorted":[],"title_canon_sha256":"dfa58ca2aad5d9cc6ae05358d18a617ce2eca118086436aca132d388a0bc2e06","abstract_canon_sha256":"dfa55a55757c8a1757a0b0700205408cc0df1ed566f735a94b84f9f40a7f1475"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:05:14.504046Z","signature_b64":"xrNkGKev4Yn1iA4fpajQpVD4SJo6TeeYNxvkKRvXHUjE1pbdD8f8opjTfQVUD3QXQGrvO3U2Wv5SxYn6y8E2Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"02d4620f6622e27f6a3c0db0f8f59ab6750aceb75fd8de4e174434e04cada5d6","last_reissued_at":"2026-07-05T08:05:14.503594Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:05:14.503594Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Investigating Regularization of Self-Play Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Abdalgader Abubaker, Mastane Achab, Mohamed El Amine Seddik, Reda Alami, Salem Lahlou","submitted_at":"2024-04-04T05:38:44Z","abstract_excerpt":"This paper explores the effects of various forms of regularization in the context of language model alignment via self-play. While both reinforcement learning from human feedback (RLHF) and direct preference optimization (DPO) require to collect costly human-annotated pairwise preferences, the self-play fine-tuning (SPIN) approach replaces the rejected answers by data generated from the previous iterate. However, the SPIN method presents a performance instability issue in the learning phase, which can be mitigated by playing against a mixture of the two previous iterates. In the same vein, we "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.04291","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/2404.04291/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":"2404.04291","created_at":"2026-07-05T08:05:14.503651+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.04291v1","created_at":"2026-07-05T08:05:14.503651+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.04291","created_at":"2026-07-05T08:05:14.503651+00:00"},{"alias_kind":"pith_short_12","alias_value":"ALKGED3GELRH","created_at":"2026-07-05T08:05:14.503651+00:00"},{"alias_kind":"pith_short_16","alias_value":"ALKGED3GELRH62R4","created_at":"2026-07-05T08:05:14.503651+00:00"},{"alias_kind":"pith_short_8","alias_value":"ALKGED3G","created_at":"2026-07-05T08:05:14.503651+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.20933","citing_title":"IRIS: Interpolative R\\'enyi Iterative Self-play for Large Language Model Fine-Tuning","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ","json":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ.json","graph_json":"https://pith.science/api/pith-number/ALKGED3GELRH62R4BWYPR5M2WZ/graph.json","events_json":"https://pith.science/api/pith-number/ALKGED3GELRH62R4BWYPR5M2WZ/events.json","paper":"https://pith.science/paper/ALKGED3G"},"agent_actions":{"view_html":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ","download_json":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ.json","view_paper":"https://pith.science/paper/ALKGED3G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.04291&json=true","fetch_graph":"https://pith.science/api/pith-number/ALKGED3GELRH62R4BWYPR5M2WZ/graph.json","fetch_events":"https://pith.science/api/pith-number/ALKGED3GELRH62R4BWYPR5M2WZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ/action/storage_attestation","attest_author":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ/action/author_attestation","sign_citation":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ/action/citation_signature","submit_replication":"https://pith.science/pith/ALKGED3GELRH62R4BWYPR5M2WZ/action/replication_record"}},"created_at":"2026-07-05T08:05:14.503651+00:00","updated_at":"2026-07-05T08:05:14.503651+00:00"}