{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5W7QFDKBCIUGKV3M53UEH34YL5","short_pith_number":"pith:5W7QFDKB","schema_version":"1.0","canonical_sha256":"edbf028d41122865576ceee843ef985f58600d46ca32c2b920f1994a29d7e6b8","source":{"kind":"arxiv","id":"2501.11463","version":2},"attestation_state":"computed","paper":{"title":"Curiosity-Driven Reinforcement Learning from Human Feedback","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"HaiFeng Wang, Haoran Sun, Hua Wu, Shuohuan Wang, Yekun Chai, Yu Sun","submitted_at":"2025-01-20T12:51:40Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) has proven effective in aligning large language models (LLMs) with human preferences, but often at the cost of reduced output diversity. This trade-off between diversity and alignment quality remains a significant challenge. Drawing inspiration from curiosity-driven exploration in reinforcement learning, we introduce curiosity-driven RLHF (CD-RLHF), a framework that incorporates intrinsic rewards for novel states, alongside traditional sparse extrinsic rewards, to optimize both output diversity and alignment quality. We demonstrate the effectiv"},"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":"2501.11463","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-01-20T12:51:40Z","cross_cats_sorted":[],"title_canon_sha256":"65c87b34a2b3129cd72a5b340d565d81c6cf5e49fae578d4c525f95cf6986bf6","abstract_canon_sha256":"264451c50ee1994d76c3ea7a5a451727631efe973f3d180cf2913f1d156e0f35"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:18.840910Z","signature_b64":"U2JEWMZcyELsb0YyOuOjfhb86F9XjpTrZr/fTyFcaQ56iE2E5d4trs7UOYtv/HRscrHFcHzw/srfIiygwRO/CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"edbf028d41122865576ceee843ef985f58600d46ca32c2b920f1994a29d7e6b8","last_reissued_at":"2026-07-05T11:13:18.840409Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:18.840409Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Curiosity-Driven Reinforcement Learning from Human Feedback","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"HaiFeng Wang, Haoran Sun, Hua Wu, Shuohuan Wang, Yekun Chai, Yu Sun","submitted_at":"2025-01-20T12:51:40Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) has proven effective in aligning large language models (LLMs) with human preferences, but often at the cost of reduced output diversity. This trade-off between diversity and alignment quality remains a significant challenge. Drawing inspiration from curiosity-driven exploration in reinforcement learning, we introduce curiosity-driven RLHF (CD-RLHF), a framework that incorporates intrinsic rewards for novel states, alongside traditional sparse extrinsic rewards, to optimize both output diversity and alignment quality. We demonstrate the effectiv"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.11463","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/2501.11463/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":"2501.11463","created_at":"2026-07-05T11:13:18.840465+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.11463v2","created_at":"2026-07-05T11:13:18.840465+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.11463","created_at":"2026-07-05T11:13:18.840465+00:00"},{"alias_kind":"pith_short_12","alias_value":"5W7QFDKBCIUG","created_at":"2026-07-05T11:13:18.840465+00:00"},{"alias_kind":"pith_short_16","alias_value":"5W7QFDKBCIUGKV3M","created_at":"2026-07-05T11:13:18.840465+00:00"},{"alias_kind":"pith_short_8","alias_value":"5W7QFDKB","created_at":"2026-07-05T11:13:18.840465+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21943","citing_title":"Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning","ref_index":187,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5","json":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5.json","graph_json":"https://pith.science/api/pith-number/5W7QFDKBCIUGKV3M53UEH34YL5/graph.json","events_json":"https://pith.science/api/pith-number/5W7QFDKBCIUGKV3M53UEH34YL5/events.json","paper":"https://pith.science/paper/5W7QFDKB"},"agent_actions":{"view_html":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5","download_json":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5.json","view_paper":"https://pith.science/paper/5W7QFDKB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.11463&json=true","fetch_graph":"https://pith.science/api/pith-number/5W7QFDKBCIUGKV3M53UEH34YL5/graph.json","fetch_events":"https://pith.science/api/pith-number/5W7QFDKBCIUGKV3M53UEH34YL5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5/action/storage_attestation","attest_author":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5/action/author_attestation","sign_citation":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5/action/citation_signature","submit_replication":"https://pith.science/pith/5W7QFDKBCIUGKV3M53UEH34YL5/action/replication_record"}},"created_at":"2026-07-05T11:13:18.840465+00:00","updated_at":"2026-07-05T11:13:18.840465+00:00"}