{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WD3VKSROPOZACPQMMQSI2WBOPK","short_pith_number":"pith:WD3VKSRO","schema_version":"1.0","canonical_sha256":"b0f7554a2e7bb2013e0c64248d582e7a90c61ec7568a628da3efb057537c434f","source":{"kind":"arxiv","id":"2606.10528","version":1},"attestation_state":"computed","paper":{"title":"Representation-Aware Advantage Estimation: Your Reward Model Provides More Than A Scalar Output","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Guozheng Li, Xiyan Fu, Yiwen Guo","submitted_at":"2026-06-09T07:57:50Z","abstract_excerpt":"Current reinforcement learning from human feedback (RLHF) methods primarily rely on scalar rewards from a trained reward model (RM). While effective, scalar rewards are often noisy and fail to capture fine-grained preference differences, whereas RM hidden states encode richer semantic and preference information. We introduce the representation-aware advantage estimation, which leverages RM hidden states and models them as auxiliary signals for better advantage estimation. Specifically, we propose the Graph-based Advantage Estimation (GraphAE), treat each sampled group as a graph, where nodes c"},"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":"2606.10528","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-09T07:57:50Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"92bd987f5b27dc98227028800b67d20a1016b0392c4124651979ac85a2f2f837","abstract_canon_sha256":"5cf04606b0ee0ffa6eed69fbec5c12a155e15c797e549216bc4337d6dad751df"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:10:24.423099Z","signature_b64":"XiAwGitcpW9C5rzKOEY31T8d2tbkks0xxLqWOg40Af/gP7Fs10tEPGdl1wwtsdlxtMD3Wubq02DeNeFgaro/AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b0f7554a2e7bb2013e0c64248d582e7a90c61ec7568a628da3efb057537c434f","last_reissued_at":"2026-06-10T01:10:24.422230Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:10:24.422230Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Representation-Aware Advantage Estimation: Your Reward Model Provides More Than A Scalar Output","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Guozheng Li, Xiyan Fu, Yiwen Guo","submitted_at":"2026-06-09T07:57:50Z","abstract_excerpt":"Current reinforcement learning from human feedback (RLHF) methods primarily rely on scalar rewards from a trained reward model (RM). While effective, scalar rewards are often noisy and fail to capture fine-grained preference differences, whereas RM hidden states encode richer semantic and preference information. We introduce the representation-aware advantage estimation, which leverages RM hidden states and models them as auxiliary signals for better advantage estimation. Specifically, we propose the Graph-based Advantage Estimation (GraphAE), treat each sampled group as a graph, where nodes c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.10528","kind":"arxiv","version":1},"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/2606.10528/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":"2606.10528","created_at":"2026-06-10T01:10:24.422379+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.10528v1","created_at":"2026-06-10T01:10:24.422379+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.10528","created_at":"2026-06-10T01:10:24.422379+00:00"},{"alias_kind":"pith_short_12","alias_value":"WD3VKSROPOZA","created_at":"2026-06-10T01:10:24.422379+00:00"},{"alias_kind":"pith_short_16","alias_value":"WD3VKSROPOZACPQM","created_at":"2026-06-10T01:10:24.422379+00:00"},{"alias_kind":"pith_short_8","alias_value":"WD3VKSRO","created_at":"2026-06-10T01:10:24.422379+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK","json":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK.json","graph_json":"https://pith.science/api/pith-number/WD3VKSROPOZACPQMMQSI2WBOPK/graph.json","events_json":"https://pith.science/api/pith-number/WD3VKSROPOZACPQMMQSI2WBOPK/events.json","paper":"https://pith.science/paper/WD3VKSRO"},"agent_actions":{"view_html":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK","download_json":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK.json","view_paper":"https://pith.science/paper/WD3VKSRO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.10528&json=true","fetch_graph":"https://pith.science/api/pith-number/WD3VKSROPOZACPQMMQSI2WBOPK/graph.json","fetch_events":"https://pith.science/api/pith-number/WD3VKSROPOZACPQMMQSI2WBOPK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK/action/storage_attestation","attest_author":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK/action/author_attestation","sign_citation":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK/action/citation_signature","submit_replication":"https://pith.science/pith/WD3VKSROPOZACPQMMQSI2WBOPK/action/replication_record"}},"created_at":"2026-06-10T01:10:24.422379+00:00","updated_at":"2026-06-10T01:10:24.422379+00:00"}