{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:YBTKZOIUQT7TVM6NCP4EB2LZE3","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3fb903e57e2c0fa4652cd732c7cf229a7045c4e1508cc2c4999bca9b681faee2","cross_cats_sorted":["cs.IT","math.IT","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-09T18:41:50Z","title_canon_sha256":"f78b627943a0cdb6f2957b1cc214519d51087f5f213826689daed2888c680a8d"},"schema_version":"1.0","source":{"id":"2406.05883","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.05883","created_at":"2026-07-05T08:29:34Z"},{"alias_kind":"arxiv_version","alias_value":"2406.05883v1","created_at":"2026-07-05T08:29:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.05883","created_at":"2026-07-05T08:29:34Z"},{"alias_kind":"pith_short_12","alias_value":"YBTKZOIUQT7T","created_at":"2026-07-05T08:29:34Z"},{"alias_kind":"pith_short_16","alias_value":"YBTKZOIUQT7TVM6N","created_at":"2026-07-05T08:29:34Z"},{"alias_kind":"pith_short_8","alias_value":"YBTKZOIU","created_at":"2026-07-05T08:29:34Z"}],"graph_snapshots":[{"event_id":"sha256:1cdf11fc067c5271ea0fc91aaae4cd03361296cc25621de85106cb3059900a26","target":"graph","created_at":"2026-07-05T08:29:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2406.05883/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Policy alignment of large language models refers to constrained policy optimization, where the policy is optimized to maximize a reward while staying close to a reference policy with respect to an $f$-divergence such as the $\\mathsf{KL}$ divergence. The best of $n$ alignment policy selects a sample from the reference policy that has the maximum reward among $n$ independent samples. For both cases (policy alignment and best of $n$), recent works showed empirically that the reward improvement of the aligned policy on the reference one scales like $\\sqrt{\\mathsf{KL}}$, with an explicit bound in $","authors_text":"Youssef Mroueh","cross_cats":["cs.IT","math.IT","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-09T18:41:50Z","title":"Information Theoretic Guarantees For Policy Alignment In Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.05883","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b0543a8b849e1d3cc387d06bd1b54689695e30b6837a4d490a67d2f90d4cd602","target":"record","created_at":"2026-07-05T08:29:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3fb903e57e2c0fa4652cd732c7cf229a7045c4e1508cc2c4999bca9b681faee2","cross_cats_sorted":["cs.IT","math.IT","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-09T18:41:50Z","title_canon_sha256":"f78b627943a0cdb6f2957b1cc214519d51087f5f213826689daed2888c680a8d"},"schema_version":"1.0","source":{"id":"2406.05883","kind":"arxiv","version":1}},"canonical_sha256":"c066acb91484ff3ab3cd13f840e97926c2c84d77efbc2fb86df6a206a8ad8266","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c066acb91484ff3ab3cd13f840e97926c2c84d77efbc2fb86df6a206a8ad8266","first_computed_at":"2026-07-05T08:29:34.922807Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:29:34.922807Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"LpGktnH0TPejM+odIrARjdVA6/TkqHO6NCNfW9fZTZj8aKy4oIk1v6faetUT+AnAmGuSU8fGel3uoIF7YybJDg==","signature_status":"signed_v1","signed_at":"2026-07-05T08:29:34.923227Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.05883","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b0543a8b849e1d3cc387d06bd1b54689695e30b6837a4d490a67d2f90d4cd602","sha256:1cdf11fc067c5271ea0fc91aaae4cd03361296cc25621de85106cb3059900a26"],"state_sha256":"6ece289d6fec5cdb7f1ca751e64c863358b68c2ebe3b359087493493aa766953"}