{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BYDZGUXSSGS4LVGKSOBRIHEJTM","short_pith_number":"pith:BYDZGUXS","schema_version":"1.0","canonical_sha256":"0e079352f291a5c5d4ca9383141c899b016c4a08c2a5adac37c0c52518dd28a5","source":{"kind":"arxiv","id":"2605.20164","version":1},"attestation_state":"computed","paper":{"title":"Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Anas Mahmoud, Bing Liu, Daniel George, Jackson Lee, MohammadHossein Rezaei, Utkarsh Tyagi, Xingang Guo, Yunzhong He","submitted_at":"2026-05-19T17:50:18Z","abstract_excerpt":"Reinforcement learning with verifiable rewards has made post-training highly effective when correctness can be checked automatically. However, many important model behaviors require satisfying several qualitative criteria at once. Rubric-based rewards address this setting by grading prompt-specific criteria and aggregating them into a scalar reward. Yet standard static aggregations conflate a criterion's human-assigned importance with its current usefulness as an optimization signal. We show that this assumption breaks down in rubric RL: many important criteria are already saturated or current"},"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":"2605.20164","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-19T17:50:18Z","cross_cats_sorted":[],"title_canon_sha256":"534d6d248c66b1f670e8640344421204f47243eee5d3e4b4b72ff4e6d685024e","abstract_canon_sha256":"db9398582f900a7e60a5f52e0f9bfbe92701857617ee8c2c862345a65d531115"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:06:05.730707Z","signature_b64":"xaPk3Rd3SvHbRLaDZaPAQs+hllL/BRvtjCfBGP4/03o+FYwpR6hnXjqHE+FkcMCoT4DLVyfetmpx1Sz1Y7sfCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0e079352f291a5c5d4ca9383141c899b016c4a08c2a5adac37c0c52518dd28a5","last_reissued_at":"2026-05-20T02:06:05.729992Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:06:05.729992Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Anas Mahmoud, Bing Liu, Daniel George, Jackson Lee, MohammadHossein Rezaei, Utkarsh Tyagi, Xingang Guo, Yunzhong He","submitted_at":"2026-05-19T17:50:18Z","abstract_excerpt":"Reinforcement learning with verifiable rewards has made post-training highly effective when correctness can be checked automatically. However, many important model behaviors require satisfying several qualitative criteria at once. Rubric-based rewards address this setting by grading prompt-specific criteria and aggregating them into a scalar reward. Yet standard static aggregations conflate a criterion's human-assigned importance with its current usefulness as an optimization signal. We show that this assumption breaks down in rubric RL: many important criteria are already saturated or current"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20164","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/2605.20164/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":"2605.20164","created_at":"2026-05-20T02:06:05.730098+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.20164v1","created_at":"2026-05-20T02:06:05.730098+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20164","created_at":"2026-05-20T02:06:05.730098+00:00"},{"alias_kind":"pith_short_12","alias_value":"BYDZGUXSSGS4","created_at":"2026-05-20T02:06:05.730098+00:00"},{"alias_kind":"pith_short_16","alias_value":"BYDZGUXSSGS4LVGK","created_at":"2026-05-20T02:06:05.730098+00:00"},{"alias_kind":"pith_short_8","alias_value":"BYDZGUXS","created_at":"2026-05-20T02:06:05.730098+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/BYDZGUXSSGS4LVGKSOBRIHEJTM","json":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM.json","graph_json":"https://pith.science/api/pith-number/BYDZGUXSSGS4LVGKSOBRIHEJTM/graph.json","events_json":"https://pith.science/api/pith-number/BYDZGUXSSGS4LVGKSOBRIHEJTM/events.json","paper":"https://pith.science/paper/BYDZGUXS"},"agent_actions":{"view_html":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM","download_json":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM.json","view_paper":"https://pith.science/paper/BYDZGUXS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.20164&json=true","fetch_graph":"https://pith.science/api/pith-number/BYDZGUXSSGS4LVGKSOBRIHEJTM/graph.json","fetch_events":"https://pith.science/api/pith-number/BYDZGUXSSGS4LVGKSOBRIHEJTM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM/action/storage_attestation","attest_author":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM/action/author_attestation","sign_citation":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM/action/citation_signature","submit_replication":"https://pith.science/pith/BYDZGUXSSGS4LVGKSOBRIHEJTM/action/replication_record"}},"created_at":"2026-05-20T02:06:05.730098+00:00","updated_at":"2026-05-20T02:06:05.730098+00:00"}