{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:XMILIN6C2UNAAKLQLQIRFLC5T5","short_pith_number":"pith:XMILIN6C","schema_version":"1.0","canonical_sha256":"bb10b437c2d51a0029705c1112ac5d9f69b283307a69c0505faeec2e0e2aab36","source":{"kind":"arxiv","id":"2309.06657","version":2},"attestation_state":"computed","paper":{"title":"Statistical Rejection Sampling Improves Preference Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jialu Liu, Misha Khalman, Mohammad Saleh, Peter J. Liu, Rishabh Joshi, Tianqi Liu, Yao Zhao","submitted_at":"2023-09-13T01:07:25Z","abstract_excerpt":"Improving the alignment of language models with human preferences remains an active research challenge. Previous approaches have primarily utilized Reinforcement Learning from Human Feedback (RLHF) via online RL methods such as Proximal Policy Optimization (PPO). Recently, offline methods such as Sequence Likelihood Calibration (SLiC) and Direct Preference Optimization (DPO) have emerged as attractive alternatives, offering improvements in stability and scalability while maintaining competitive performance. SLiC refines its loss function using sequence pairs sampled from a supervised fine-tune"},"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":"2309.06657","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-13T01:07:25Z","cross_cats_sorted":[],"title_canon_sha256":"94c371d54ceabb0582b8bcfdf67f02e09b93b9e8698f1b426d516bc61f001cee","abstract_canon_sha256":"9dc78268a381bfb9e94936621d5d1120d76e61a58f43adc2e52b421185d705c9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:36:54.159259Z","signature_b64":"AvRwHb1iv70KaMCe2skUgZfrCY9eSDPZ99HCxwiqABPSozMdo9vL/KzprPc0f5tx4fFUfGnW0rw+Y6dhSPhiBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bb10b437c2d51a0029705c1112ac5d9f69b283307a69c0505faeec2e0e2aab36","last_reissued_at":"2026-07-05T07:36:54.158850Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:36:54.158850Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Statistical Rejection Sampling Improves Preference Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jialu Liu, Misha Khalman, Mohammad Saleh, Peter J. Liu, Rishabh Joshi, Tianqi Liu, Yao Zhao","submitted_at":"2023-09-13T01:07:25Z","abstract_excerpt":"Improving the alignment of language models with human preferences remains an active research challenge. Previous approaches have primarily utilized Reinforcement Learning from Human Feedback (RLHF) via online RL methods such as Proximal Policy Optimization (PPO). Recently, offline methods such as Sequence Likelihood Calibration (SLiC) and Direct Preference Optimization (DPO) have emerged as attractive alternatives, offering improvements in stability and scalability while maintaining competitive performance. SLiC refines its loss function using sequence pairs sampled from a supervised fine-tune"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.06657","kind":"arxiv","version":2},"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/2309.06657/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":"2309.06657","created_at":"2026-07-05T07:36:54.158910+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.06657v2","created_at":"2026-07-05T07:36:54.158910+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.06657","created_at":"2026-07-05T07:36:54.158910+00:00"},{"alias_kind":"pith_short_12","alias_value":"XMILIN6C2UNA","created_at":"2026-07-05T07:36:54.158910+00:00"},{"alias_kind":"pith_short_16","alias_value":"XMILIN6C2UNAAKLQ","created_at":"2026-07-05T07:36:54.158910+00:00"},{"alias_kind":"pith_short_8","alias_value":"XMILIN6C","created_at":"2026-07-05T07:36:54.158910+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09635","citing_title":"Gradient-Guided Reward Optimization for Inference-time Alignment","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28707","citing_title":"BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewards","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2510.07972","citing_title":"SHE: Stepwise Hybrid Examination Reinforcement Learning Framework for E-commerce Search Relevance","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2411.10442","citing_title":"Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":112,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":112,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06650","citing_title":"Beyond Negative Rollouts: Positive-Only Policy Optimization with Implicit Negative Gradients","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18946","citing_title":"Reasoning Structure Matters for Safety Alignment of Reasoning Models","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06987","citing_title":"Response Time Enhances Alignment with Heterogeneous Preferences","ref_index":166,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14895","citing_title":"Beyond Importance Sampling: Rejection-Gated Policy Optimization","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20727","citing_title":"Supplement Generation Training for Enhancing Agentic Task Performance","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20140","citing_title":"HiPO: Hierarchical Preference Optimization for Adaptive Reasoning in LLMs","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02495","citing_title":"Efficient Preference Poisoning Attack on Offline RLHF","ref_index":80,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5","json":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5.json","graph_json":"https://pith.science/api/pith-number/XMILIN6C2UNAAKLQLQIRFLC5T5/graph.json","events_json":"https://pith.science/api/pith-number/XMILIN6C2UNAAKLQLQIRFLC5T5/events.json","paper":"https://pith.science/paper/XMILIN6C"},"agent_actions":{"view_html":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5","download_json":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5.json","view_paper":"https://pith.science/paper/XMILIN6C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.06657&json=true","fetch_graph":"https://pith.science/api/pith-number/XMILIN6C2UNAAKLQLQIRFLC5T5/graph.json","fetch_events":"https://pith.science/api/pith-number/XMILIN6C2UNAAKLQLQIRFLC5T5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5/action/storage_attestation","attest_author":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5/action/author_attestation","sign_citation":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5/action/citation_signature","submit_replication":"https://pith.science/pith/XMILIN6C2UNAAKLQLQIRFLC5T5/action/replication_record"}},"created_at":"2026-07-05T07:36:54.158910+00:00","updated_at":"2026-07-05T07:36:54.158910+00:00"}