{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:574LWMSMGCNB5ZVRVGRPQ65D25","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":"e3a1e8110a18ab3ba1ef4c03967d65cae0732e8299d33dda91259a683fabc5a8","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-23T22:00:38Z","title_canon_sha256":"b663e3c6535e6b8f5e2f8b667cca5fdf070daefd5a8c0a0ba0fb65e1593c9d1a"},"schema_version":"1.0","source":{"id":"2605.24749","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.24749","created_at":"2026-05-26T01:03:56Z"},{"alias_kind":"arxiv_version","alias_value":"2605.24749v1","created_at":"2026-05-26T01:03:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24749","created_at":"2026-05-26T01:03:56Z"},{"alias_kind":"pith_short_12","alias_value":"574LWMSMGCNB","created_at":"2026-05-26T01:03:56Z"},{"alias_kind":"pith_short_16","alias_value":"574LWMSMGCNB5ZVR","created_at":"2026-05-26T01:03:56Z"},{"alias_kind":"pith_short_8","alias_value":"574LWMSM","created_at":"2026-05-26T01:03:56Z"}],"graph_snapshots":[{"event_id":"sha256:31671ec0124b56ad44702f1b1626db59a3c335c3ceb4fec8b13fc21666a646c2","target":"graph","created_at":"2026-05-26T01:03:56Z","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/2605.24749/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reward modeling is not only a prediction problem: in KL-regularized policy optimization, the learned reward is exponentiated to define the deployed policy, so downstream value depends on errors in reward-tilted regions. We study this feedback in a Gaussian single-index model with $r^*(x) = \\sigma^*(\\langle \\theta^*, x\\rangle)$ and $x \\sim N(0, I_d)$. We analyze a two-stage neural reward model that first learns the hidden direction $\\theta^*$ from reward-weighted samples and then fits the readout layer by weighted ridge regression. Exponential reward weighting changes the Hermite signal availab","authors_text":"Akifumi Wachi, Kohei Miyaguchi, Rei Higuchi, Ryotaro Kawata, Shokichi Takakura, Taiji Suzuki","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-23T22:00:38Z","title":"How Neural Reward Models Learn Features for Policy Optimization: A Single-Index Analysis"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24749","kind":"arxiv","version":1},"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:eaed9c88e82664ad6f89089aefb4e3e4cc6f791619b65dc05daa6d8ffdaf1a86","target":"record","created_at":"2026-05-26T01:03:56Z","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":"e3a1e8110a18ab3ba1ef4c03967d65cae0732e8299d33dda91259a683fabc5a8","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-23T22:00:38Z","title_canon_sha256":"b663e3c6535e6b8f5e2f8b667cca5fdf070daefd5a8c0a0ba0fb65e1593c9d1a"},"schema_version":"1.0","source":{"id":"2605.24749","kind":"arxiv","version":1}},"canonical_sha256":"eff8bb324c309a1ee6b1a9a2f87ba3d7662f6f91a89dd256ec1694b80a53c4e4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eff8bb324c309a1ee6b1a9a2f87ba3d7662f6f91a89dd256ec1694b80a53c4e4","first_computed_at":"2026-05-26T01:03:56.317717Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T01:03:56.317717Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"feNsUihYSNdCqf4BH7S4ZdzGK+ggzLjzkkcLoRWXRR5lrx/+/IJOEWWxeyr+ca8EPzVxpPDtKtGUzTPByrcRAw==","signature_status":"signed_v1","signed_at":"2026-05-26T01:03:56.318506Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.24749","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:eaed9c88e82664ad6f89089aefb4e3e4cc6f791619b65dc05daa6d8ffdaf1a86","sha256:31671ec0124b56ad44702f1b1626db59a3c335c3ceb4fec8b13fc21666a646c2"],"state_sha256":"56d4d0233f29eb60236aeda73c5e816bc630c54bee28a4ee659b7bca4c7b7f5f"}