{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:XHED7ZH4CVCKIOPSDCL4BT5NJV","short_pith_number":"pith:XHED7ZH4","schema_version":"1.0","canonical_sha256":"b9c83fe4fc1544a439f21897c0cfad4d716d2c31f923c2d3715de71dc0325558","source":{"kind":"arxiv","id":"2306.01693","version":2},"attestation_state":"computed","paper":{"title":"Fine-Grained Human Feedback Gives Better Rewards for Language Model Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alane Suhr, Hannaneh Hajishirzi, Mari Ostendorf, Noah A. Smith, Nouha Dziri, Prithviraj Ammanabrolu, Weijia Shi, Yushi Hu, Zeqiu Wu","submitted_at":"2023-06-02T17:11:37Z","abstract_excerpt":"Language models (LMs) often exhibit undesirable text generation behaviors, including generating false, toxic, or irrelevant outputs. Reinforcement learning from human feedback (RLHF) - where human preference judgments on LM outputs are transformed into a learning signal - has recently shown promise in addressing these issues. However, such holistic feedback conveys limited information on long text outputs; it does not indicate which aspects of the outputs influenced user preference; e.g., which parts contain what type(s) of errors. In this paper, we use fine-grained human feedback (e.g., which"},"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":"2306.01693","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-06-02T17:11:37Z","cross_cats_sorted":[],"title_canon_sha256":"2af15a2d0e8ab8c746885e80d4d81c0aeed32c07892d205b00226f1dba36caed","abstract_canon_sha256":"1a4c5877b5736f2a708e99e01f080a8fcf2ee6077e62de62349d5c91059bb718"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:06:29.171094Z","signature_b64":"U0Y6AFh72y9OfjyqyO7YLixa8IQkP9SxjZlPVhgLjkBR6Dl4FN1ou/Kmj92pdhE1KuqT+06iC2h5gGb6bIEZBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9c83fe4fc1544a439f21897c0cfad4d716d2c31f923c2d3715de71dc0325558","last_reissued_at":"2026-07-05T07:06:29.170613Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:06:29.170613Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-Grained Human Feedback Gives Better Rewards for Language Model Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alane Suhr, Hannaneh Hajishirzi, Mari Ostendorf, Noah A. Smith, Nouha Dziri, Prithviraj Ammanabrolu, Weijia Shi, Yushi Hu, Zeqiu Wu","submitted_at":"2023-06-02T17:11:37Z","abstract_excerpt":"Language models (LMs) often exhibit undesirable text generation behaviors, including generating false, toxic, or irrelevant outputs. Reinforcement learning from human feedback (RLHF) - where human preference judgments on LM outputs are transformed into a learning signal - has recently shown promise in addressing these issues. However, such holistic feedback conveys limited information on long text outputs; it does not indicate which aspects of the outputs influenced user preference; e.g., which parts contain what type(s) of errors. In this paper, we use fine-grained human feedback (e.g., which"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.01693","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/2306.01693/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":"2306.01693","created_at":"2026-07-05T07:06:29.170672+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.01693v2","created_at":"2026-07-05T07:06:29.170672+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.01693","created_at":"2026-07-05T07:06:29.170672+00:00"},{"alias_kind":"pith_short_12","alias_value":"XHED7ZH4CVCK","created_at":"2026-07-05T07:06:29.170672+00:00"},{"alias_kind":"pith_short_16","alias_value":"XHED7ZH4CVCKIOPS","created_at":"2026-07-05T07:06:29.170672+00:00"},{"alias_kind":"pith_short_8","alias_value":"XHED7ZH4","created_at":"2026-07-05T07:06:29.170672+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06674","citing_title":"What Do People Actually Want From AI? Mapping Preference Plurality","ref_index":90,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20506","citing_title":"Reinforcing Human Behavior Simulation via Verbal Feedback","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15300","citing_title":"Deep Pre-Alignment for VLMs","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2509.11206","citing_title":"Evalet: Evaluating Large Language Models through Functional Fragmentation","ref_index":93,"is_internal_anchor":false},{"citing_arxiv_id":"2309.11495","citing_title":"Chain-of-Verification Reduces Hallucination in Large Language Models","ref_index":116,"is_internal_anchor":false},{"citing_arxiv_id":"2403.17297","citing_title":"InternLM2 Technical Report","ref_index":166,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":135,"is_internal_anchor":false},{"citing_arxiv_id":"2312.08935","citing_title":"Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":135,"is_internal_anchor":false},{"citing_arxiv_id":"2310.11511","citing_title":"Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection","ref_index":127,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV","json":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV.json","graph_json":"https://pith.science/api/pith-number/XHED7ZH4CVCKIOPSDCL4BT5NJV/graph.json","events_json":"https://pith.science/api/pith-number/XHED7ZH4CVCKIOPSDCL4BT5NJV/events.json","paper":"https://pith.science/paper/XHED7ZH4"},"agent_actions":{"view_html":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV","download_json":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV.json","view_paper":"https://pith.science/paper/XHED7ZH4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.01693&json=true","fetch_graph":"https://pith.science/api/pith-number/XHED7ZH4CVCKIOPSDCL4BT5NJV/graph.json","fetch_events":"https://pith.science/api/pith-number/XHED7ZH4CVCKIOPSDCL4BT5NJV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV/action/storage_attestation","attest_author":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV/action/author_attestation","sign_citation":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV/action/citation_signature","submit_replication":"https://pith.science/pith/XHED7ZH4CVCKIOPSDCL4BT5NJV/action/replication_record"}},"created_at":"2026-07-05T07:06:29.170672+00:00","updated_at":"2026-07-05T07:06:29.170672+00:00"}