{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LPC2C36SAWR3V377CU6DT2TP2C","short_pith_number":"pith:LPC2C36S","schema_version":"1.0","canonical_sha256":"5bc5a16fd205a3baefff153c39ea6fd09ff743a99953c2b2c5e47ce4e7bf2d44","source":{"kind":"arxiv","id":"2407.02119","version":2},"attestation_state":"computed","paper":{"title":"Cost-Effective Proxy Reward Model Construction with On-Policy and Active Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Donghan Yu, Hiteshi Sharma, Kevin Jamieson, Nikos Karampatziakis, Shuohang Wang, Simon Shaolei Du, Yelong Shen, Yifang Chen, Ziyi Yang","submitted_at":"2024-07-02T10:09:19Z","abstract_excerpt":"Reinforcement learning with human feedback (RLHF), as a widely adopted approach in current large language model pipelines, is \\textit{bottlenecked by the size of human preference data}. While traditional methods rely on offline preference dataset constructions, recent approaches have shifted towards online settings, where a learner uses a small amount of labeled seed data and a large pool of unlabeled prompts to iteratively construct new preference data through self-generated responses and high-quality reward/preference feedback. However, most current online algorithms still focus on preferenc"},"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":"2407.02119","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-07-02T10:09:19Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"61d7b89cc099c0134abd82495166b19f79f20b741bbb5391d9b6b47e40f0a353","abstract_canon_sha256":"65c09cb7d0ec82b5884f06242f42a470b589fcaab4d23ac184898009ed6ea8ef"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:41:46.459701Z","signature_b64":"rll8BJVZIjsk3BKqseBBbpf5RJ/jhvBVON94DNiFMegPTLbHzdrjYEFYzNrV/sngYHwQxoqxzpw2uFvzpRIGDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5bc5a16fd205a3baefff153c39ea6fd09ff743a99953c2b2c5e47ce4e7bf2d44","last_reissued_at":"2026-07-05T08:41:46.459260Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:41:46.459260Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cost-Effective Proxy Reward Model Construction with On-Policy and Active Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Donghan Yu, Hiteshi Sharma, Kevin Jamieson, Nikos Karampatziakis, Shuohang Wang, Simon Shaolei Du, Yelong Shen, Yifang Chen, Ziyi Yang","submitted_at":"2024-07-02T10:09:19Z","abstract_excerpt":"Reinforcement learning with human feedback (RLHF), as a widely adopted approach in current large language model pipelines, is \\textit{bottlenecked by the size of human preference data}. While traditional methods rely on offline preference dataset constructions, recent approaches have shifted towards online settings, where a learner uses a small amount of labeled seed data and a large pool of unlabeled prompts to iteratively construct new preference data through self-generated responses and high-quality reward/preference feedback. However, most current online algorithms still focus on preferenc"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.02119","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/2407.02119/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":"2407.02119","created_at":"2026-07-05T08:41:46.459323+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.02119v2","created_at":"2026-07-05T08:41:46.459323+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.02119","created_at":"2026-07-05T08:41:46.459323+00:00"},{"alias_kind":"pith_short_12","alias_value":"LPC2C36SAWR3","created_at":"2026-07-05T08:41:46.459323+00:00"},{"alias_kind":"pith_short_16","alias_value":"LPC2C36SAWR3V377","created_at":"2026-07-05T08:41:46.459323+00:00"},{"alias_kind":"pith_short_8","alias_value":"LPC2C36S","created_at":"2026-07-05T08:41:46.459323+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11859","citing_title":"EvoNav: Evolutionary Reward Function Design for Robot Navigation with Large Language Models","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C","json":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C.json","graph_json":"https://pith.science/api/pith-number/LPC2C36SAWR3V377CU6DT2TP2C/graph.json","events_json":"https://pith.science/api/pith-number/LPC2C36SAWR3V377CU6DT2TP2C/events.json","paper":"https://pith.science/paper/LPC2C36S"},"agent_actions":{"view_html":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C","download_json":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C.json","view_paper":"https://pith.science/paper/LPC2C36S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.02119&json=true","fetch_graph":"https://pith.science/api/pith-number/LPC2C36SAWR3V377CU6DT2TP2C/graph.json","fetch_events":"https://pith.science/api/pith-number/LPC2C36SAWR3V377CU6DT2TP2C/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C/action/storage_attestation","attest_author":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C/action/author_attestation","sign_citation":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C/action/citation_signature","submit_replication":"https://pith.science/pith/LPC2C36SAWR3V377CU6DT2TP2C/action/replication_record"}},"created_at":"2026-07-05T08:41:46.459323+00:00","updated_at":"2026-07-05T08:41:46.459323+00:00"}