{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ONM3W2MH2FZO2JWGI4BM2KVYHH","short_pith_number":"pith:ONM3W2MH","schema_version":"1.0","canonical_sha256":"7359bb6987d172ed26c64702cd2ab839dbe328bb74e4f56c07f667c2fa13910d","source":{"kind":"arxiv","id":"2403.13787","version":2},"attestation_state":"computed","paper":{"title":"RewardBench: Evaluating Reward Models for Language Modeling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bill Yuchen Lin, Hannaneh Hajishirzi, Jacob Morrison, Khyathi Chandu, LJ Miranda, Nathan Lambert, Noah A. Smith, Nouha Dziri, Sachin Kumar, Tom Zick, Valentina Pyatkin, Yejin Choi","submitted_at":"2024-03-20T17:49:54Z","abstract_excerpt":"Reward models (RMs) are at the crux of successfully using RLHF to align pretrained models to human preferences, yet there has been relatively little study that focuses on evaluation of those models. Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models and which values are embedded in them. Resources for reward model training and understanding are sparse in the nascent open-source community around them. To enhance scientific understanding of reward models, we present RewardBench, a benchmark dataset and code-base for evalua"},"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":"2403.13787","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-03-20T17:49:54Z","cross_cats_sorted":[],"title_canon_sha256":"1bbf49fa9ac42d7fd478b8f6cb538064aa8ec7952d34e3da5bf7fe943e3f3d4f","abstract_canon_sha256":"1c0024b059f8647810fd6c17e064e7b5baa2a8c4e417075655203c1cbbcfed67"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:29:24.623485Z","signature_b64":"oIEXGv9jf71yxUrWzgPcLssK4M0g0BjYrvSoGRhNVIgYAFCgVwt5DeRtMebCCyHHY7OaToC+CgO+hKNclByfAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7359bb6987d172ed26c64702cd2ab839dbe328bb74e4f56c07f667c2fa13910d","last_reissued_at":"2026-07-05T08:29:24.622998Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:29:24.622998Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RewardBench: Evaluating Reward Models for Language Modeling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bill Yuchen Lin, Hannaneh Hajishirzi, Jacob Morrison, Khyathi Chandu, LJ Miranda, Nathan Lambert, Noah A. Smith, Nouha Dziri, Sachin Kumar, Tom Zick, Valentina Pyatkin, Yejin Choi","submitted_at":"2024-03-20T17:49:54Z","abstract_excerpt":"Reward models (RMs) are at the crux of successfully using RLHF to align pretrained models to human preferences, yet there has been relatively little study that focuses on evaluation of those models. Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models and which values are embedded in them. Resources for reward model training and understanding are sparse in the nascent open-source community around them. To enhance scientific understanding of reward models, we present RewardBench, a benchmark dataset and code-base for evalua"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.13787","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/2403.13787/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":"2403.13787","created_at":"2026-07-05T08:29:24.623053+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.13787v2","created_at":"2026-07-05T08:29:24.623053+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.13787","created_at":"2026-07-05T08:29:24.623053+00:00"},{"alias_kind":"pith_short_12","alias_value":"ONM3W2MH2FZO","created_at":"2026-07-05T08:29:24.623053+00:00"},{"alias_kind":"pith_short_16","alias_value":"ONM3W2MH2FZO2JWG","created_at":"2026-07-05T08:29:24.623053+00:00"},{"alias_kind":"pith_short_8","alias_value":"ONM3W2MH","created_at":"2026-07-05T08:29:24.623053+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":30,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09711","citing_title":"Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01034","citing_title":"A Finite-Calibration Regime Map for LLM Judge Panels","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30803","citing_title":"PReMISE: Policy Rubrics as Measurement Specifications for LLM Judges","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30758","citing_title":"Pairwise Reference Alignment as a Model-Level Ordinal Observable","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31748","citing_title":"Addressing Over-Refusal in LLMs with Competing Rewards","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2606.32034","citing_title":"QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00437","citing_title":"EST-PRM: Stress-Testing Process Reward Models Before They Become Load-Bearing","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21296","citing_title":"Discriminatory Compliance: How LLMs Answer Queries from Protected Groups","ref_index":97,"is_internal_anchor":false},{"citing_arxiv_id":"2412.15115","citing_title":"Qwen2.5 Technical Report","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2508.04149","citing_title":"Difficulty-Based Preference Data Selection by DPO Implicit Reward Gap","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16339","citing_title":"Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20164","citing_title":"Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2506.01937","citing_title":"RewardBench 2: Advancing Reward Model Evaluation","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12382","citing_title":"Exploring the Secondary Risks of Large Language Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2506.16791","citing_title":"TabArena: A Living Benchmark for Machine Learning on Tabular Data","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2509.11206","citing_title":"Evalet: Evaluating Large Language Models through Functional Fragmentation","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04225","citing_title":"Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2510.24235","citing_title":"PaTaRM: Bridging Pairwise and Pointwise Signals via Preference-Aware Task-Adaptive Reward Modeling","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2410.18451","citing_title":"Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2406.08464","citing_title":"Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2501.13918","citing_title":"Improving Video Generation with Human Feedback","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2412.05579","citing_title":"LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods","ref_index":120,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23178","citing_title":"Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19457","citing_title":"Four-Axis Decision Alignment for Long-Horizon Enterprise AI Agents","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2511.16624","citing_title":"SAM 3D: 3Dfy Anything in Images","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH","json":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH.json","graph_json":"https://pith.science/api/pith-number/ONM3W2MH2FZO2JWGI4BM2KVYHH/graph.json","events_json":"https://pith.science/api/pith-number/ONM3W2MH2FZO2JWGI4BM2KVYHH/events.json","paper":"https://pith.science/paper/ONM3W2MH"},"agent_actions":{"view_html":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH","download_json":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH.json","view_paper":"https://pith.science/paper/ONM3W2MH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.13787&json=true","fetch_graph":"https://pith.science/api/pith-number/ONM3W2MH2FZO2JWGI4BM2KVYHH/graph.json","fetch_events":"https://pith.science/api/pith-number/ONM3W2MH2FZO2JWGI4BM2KVYHH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH/action/storage_attestation","attest_author":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH/action/author_attestation","sign_citation":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH/action/citation_signature","submit_replication":"https://pith.science/pith/ONM3W2MH2FZO2JWGI4BM2KVYHH/action/replication_record"}},"created_at":"2026-07-05T08:29:24.623053+00:00","updated_at":"2026-07-05T08:29:24.623053+00:00"}