{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:3M75WMYOFCPZWBFN26H7JWVAGS","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"611f09584a763fd3d4aa47df4740f36a2158dba020feb4797fb9c7ef8225bdd9","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-18T21:38:21Z","title_canon_sha256":"b737530287134d059cab2328ba8c6d6f04742cd2c73b92b2f3def4e33c1d9f73"},"schema_version":"1.0","source":{"id":"2410.14872","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.14872","created_at":"2026-07-05T09:24:18Z"},{"alias_kind":"arxiv_version","alias_value":"2410.14872v2","created_at":"2026-07-05T09:24:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.14872","created_at":"2026-07-05T09:24:18Z"},{"alias_kind":"pith_short_12","alias_value":"3M75WMYOFCPZ","created_at":"2026-07-05T09:24:18Z"},{"alias_kind":"pith_short_16","alias_value":"3M75WMYOFCPZWBFN","created_at":"2026-07-05T09:24:18Z"},{"alias_kind":"pith_short_8","alias_value":"3M75WMYO","created_at":"2026-07-05T09:24:18Z"}],"graph_snapshots":[{"event_id":"sha256:b953ed83a39ed42d85e6442b1650d38429bdae9e0c810a504f07d5215faa62f0","target":"graph","created_at":"2026-07-05T09:24:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2410.14872/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We introduce a new benchmark for reward models that quantifies their ability to produce strong language models through RLHF (Reinforcement Learning from Human Feedback). The gold-standard approach is to run a full RLHF training pipeline and directly probe downstream LLM performance. However, this process is prohibitively expensive. To address this, we build a predictive model of downstream LLM performance by evaluating the reward model on proxy tasks. These proxy tasks consist of a large-scale human preference and a verifiable correctness preference dataset, in which we measure 12 metrics acro","authors_text":"Anastasios N. Angelopoulos, Banghua Zhu, Connor Chen, Evan Frick, Ion Stoica, Jiantao Jiao, Joseph E. Gonzalez, Tianle Li, Wei-Lin Chiang","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-18T21:38:21Z","title":"How to Evaluate Reward Models for RLHF"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.14872","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:332aecfe4f2b539a2e27d66c841619f8f8f358ca060ba2393ac0989ec96f77d2","target":"record","created_at":"2026-07-05T09:24:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"611f09584a763fd3d4aa47df4740f36a2158dba020feb4797fb9c7ef8225bdd9","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-18T21:38:21Z","title_canon_sha256":"b737530287134d059cab2328ba8c6d6f04742cd2c73b92b2f3def4e33c1d9f73"},"schema_version":"1.0","source":{"id":"2410.14872","kind":"arxiv","version":2}},"canonical_sha256":"db3fdb330e289f9b04add78ff4daa034929997a973031a6013ed9ca1f92c4340","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"db3fdb330e289f9b04add78ff4daa034929997a973031a6013ed9ca1f92c4340","first_computed_at":"2026-07-05T09:24:18.805292Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:24:18.805292Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"+i7eTlT6LPaShlp0baDwov8qy34tSmxKcxJBllabSra62l4ZoHxW/trqp2xAuEmeYll5f6iMDlO8vDhp3kh8CQ==","signature_status":"signed_v1","signed_at":"2026-07-05T09:24:18.805808Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.14872","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:332aecfe4f2b539a2e27d66c841619f8f8f358ca060ba2393ac0989ec96f77d2","sha256:b953ed83a39ed42d85e6442b1650d38429bdae9e0c810a504f07d5215faa62f0"],"state_sha256":"d787c7b1ca811b4dccfe6d1134d27ebaee197cc9232c9ac1260d64fdab422a8c"}