{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:X4LXBMEXMK5IMIHMD4DZRPUUZL","short_pith_number":"pith:X4LXBMEX","schema_version":"1.0","canonical_sha256":"bf1770b09762ba8620ec1f0798be94cacd3455e261125568585dc793664dcaa7","source":{"kind":"arxiv","id":"2606.00424","version":1},"attestation_state":"computed","paper":{"title":"Weak Critics Make Strong Learners: On-Policy Critique Distillation for Scalable Oversight","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Can Jin, Dimitris N. Metaxas, Eddy Zhang, Jiakang Li, Rui Wu","submitted_at":"2026-05-29T23:21:48Z","abstract_excerpt":"As large language models become stronger, weak supervisors may fail to provide reliable labels, preferences, or final judgments for complex outputs, limiting both weak-to-strong generalization and scalable oversight. We study a more tractable form of weak supervision: using a weak model as a critic rather than as a labeler or judge. Instead of solving the task or selecting the correct answer, the weak critic only needs to provide a non-misleading revision direction that helps the strong model better use its own knowledge. We call this setting *weak-critic strong oversight*. We first show that "},"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":"2606.00424","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-29T23:21:48Z","cross_cats_sorted":[],"title_canon_sha256":"98436ca50737373f203f4df95a31d52e6fc727123f1a66f8b4e4e4b1e378e312","abstract_canon_sha256":"62aeb66618cae50da374caa0ebfb87ebd1fd7202402e27f8a7b255cb47d47d3f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:54.239386Z","signature_b64":"bYLpNcQy++wisW4aa4ya15W1b4uV744q3iAT3HBUNRS5zpNRo5nQWmuAA9R6JAmy4S31rYAj1w0IwfqTqStICg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bf1770b09762ba8620ec1f0798be94cacd3455e261125568585dc793664dcaa7","last_reissued_at":"2026-06-02T01:03:54.238980Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:54.238980Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Weak Critics Make Strong Learners: On-Policy Critique Distillation for Scalable Oversight","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Can Jin, Dimitris N. Metaxas, Eddy Zhang, Jiakang Li, Rui Wu","submitted_at":"2026-05-29T23:21:48Z","abstract_excerpt":"As large language models become stronger, weak supervisors may fail to provide reliable labels, preferences, or final judgments for complex outputs, limiting both weak-to-strong generalization and scalable oversight. We study a more tractable form of weak supervision: using a weak model as a critic rather than as a labeler or judge. Instead of solving the task or selecting the correct answer, the weak critic only needs to provide a non-misleading revision direction that helps the strong model better use its own knowledge. We call this setting *weak-critic strong oversight*. We first show that "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00424","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/2606.00424/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":"2606.00424","created_at":"2026-06-02T01:03:54.239035+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00424v1","created_at":"2026-06-02T01:03:54.239035+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00424","created_at":"2026-06-02T01:03:54.239035+00:00"},{"alias_kind":"pith_short_12","alias_value":"X4LXBMEXMK5I","created_at":"2026-06-02T01:03:54.239035+00:00"},{"alias_kind":"pith_short_16","alias_value":"X4LXBMEXMK5IMIHM","created_at":"2026-06-02T01:03:54.239035+00:00"},{"alias_kind":"pith_short_8","alias_value":"X4LXBMEX","created_at":"2026-06-02T01:03:54.239035+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/X4LXBMEXMK5IMIHMD4DZRPUUZL","json":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL.json","graph_json":"https://pith.science/api/pith-number/X4LXBMEXMK5IMIHMD4DZRPUUZL/graph.json","events_json":"https://pith.science/api/pith-number/X4LXBMEXMK5IMIHMD4DZRPUUZL/events.json","paper":"https://pith.science/paper/X4LXBMEX"},"agent_actions":{"view_html":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL","download_json":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL.json","view_paper":"https://pith.science/paper/X4LXBMEX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00424&json=true","fetch_graph":"https://pith.science/api/pith-number/X4LXBMEXMK5IMIHMD4DZRPUUZL/graph.json","fetch_events":"https://pith.science/api/pith-number/X4LXBMEXMK5IMIHMD4DZRPUUZL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL/action/storage_attestation","attest_author":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL/action/author_attestation","sign_citation":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL/action/citation_signature","submit_replication":"https://pith.science/pith/X4LXBMEXMK5IMIHMD4DZRPUUZL/action/replication_record"}},"created_at":"2026-06-02T01:03:54.239035+00:00","updated_at":"2026-06-02T01:03:54.239035+00:00"}