{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XI2ULD5AVL5AT3AVIGWQIIOG65","short_pith_number":"pith:XI2ULD5A","schema_version":"1.0","canonical_sha256":"ba35458fa0aafa09ec1541ad0421c6f74da072d4209fccd4b01bdb38907cf7e2","source":{"kind":"arxiv","id":"2606.21811","version":1},"attestation_state":"computed","paper":{"title":"Steer, Don't Solve: Training Small Critic Models for Large Code Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.SE","authors_text":"Atharva Naik, Carolyn Rose, Ruichen Zhu, Shubham Gandhi, Yiqing Xie","submitted_at":"2026-06-20T00:14:38Z","abstract_excerpt":"End-to-end code agent training is resource-intensive and plateaus on the strategy-level reasoning needed to resolve code issues, since jointly optimizing code-level execution and strategy-level reasoning leaves the latter underdeveloped. Instead, we freeze the agent and add a critic model to supply that signal. Prior code critics are post-hoc, scoring completed trajectories rather than steering the agent; we instead train a small critic that provides intra-trajectory feedback via Supervised Fine-Tuning. On SWE-bench Verified, a critic trained on CWM-32B trajectories transfers to two unseen age"},"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":"2606.21811","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-06-20T00:14:38Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"4937f5bc5dbabc252e9165c6bf26dac9d172a53e8c968b247b9dd926fa838b45","abstract_canon_sha256":"faa5eee88df4601ce191b5e95527cd735fb755f672eb50287d6f4bd64b6226dc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T01:13:23.430643Z","signature_b64":"b+6YxykQOjqM8YiFz867Y4MSLpA4enr4gYFzsQlu11yd8iIAeoiJYgcZcZa2pfuugmTeZ2Xq3WOzTXwctsYgDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ba35458fa0aafa09ec1541ad0421c6f74da072d4209fccd4b01bdb38907cf7e2","last_reissued_at":"2026-06-23T01:13:23.429993Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T01:13:23.429993Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Steer, Don't Solve: Training Small Critic Models for Large Code Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.SE","authors_text":"Atharva Naik, Carolyn Rose, Ruichen Zhu, Shubham Gandhi, Yiqing Xie","submitted_at":"2026-06-20T00:14:38Z","abstract_excerpt":"End-to-end code agent training is resource-intensive and plateaus on the strategy-level reasoning needed to resolve code issues, since jointly optimizing code-level execution and strategy-level reasoning leaves the latter underdeveloped. Instead, we freeze the agent and add a critic model to supply that signal. Prior code critics are post-hoc, scoring completed trajectories rather than steering the agent; we instead train a small critic that provides intra-trajectory feedback via Supervised Fine-Tuning. On SWE-bench Verified, a critic trained on CWM-32B trajectories transfers to two unseen age"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.21811","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/2606.21811/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":"2606.21811","created_at":"2026-06-23T01:13:23.430078+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.21811v1","created_at":"2026-06-23T01:13:23.430078+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.21811","created_at":"2026-06-23T01:13:23.430078+00:00"},{"alias_kind":"pith_short_12","alias_value":"XI2ULD5AVL5A","created_at":"2026-06-23T01:13:23.430078+00:00"},{"alias_kind":"pith_short_16","alias_value":"XI2ULD5AVL5AT3AV","created_at":"2026-06-23T01:13:23.430078+00:00"},{"alias_kind":"pith_short_8","alias_value":"XI2ULD5A","created_at":"2026-06-23T01:13:23.430078+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/XI2ULD5AVL5AT3AVIGWQIIOG65","json":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65.json","graph_json":"https://pith.science/api/pith-number/XI2ULD5AVL5AT3AVIGWQIIOG65/graph.json","events_json":"https://pith.science/api/pith-number/XI2ULD5AVL5AT3AVIGWQIIOG65/events.json","paper":"https://pith.science/paper/XI2ULD5A"},"agent_actions":{"view_html":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65","download_json":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65.json","view_paper":"https://pith.science/paper/XI2ULD5A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.21811&json=true","fetch_graph":"https://pith.science/api/pith-number/XI2ULD5AVL5AT3AVIGWQIIOG65/graph.json","fetch_events":"https://pith.science/api/pith-number/XI2ULD5AVL5AT3AVIGWQIIOG65/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65/action/storage_attestation","attest_author":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65/action/author_attestation","sign_citation":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65/action/citation_signature","submit_replication":"https://pith.science/pith/XI2ULD5AVL5AT3AVIGWQIIOG65/action/replication_record"}},"created_at":"2026-06-23T01:13:23.430078+00:00","updated_at":"2026-06-23T01:13:23.430078+00:00"}