{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ACB6IZPAQU6P7URY5VC7M7JUZB","short_pith_number":"pith:ACB6IZPA","schema_version":"1.0","canonical_sha256":"0083e465e0853cffd238ed45f67d34c85fa7db335b9bc25cfa6859741cefe009","source":{"kind":"arxiv","id":"2606.28438","version":1},"attestation_state":"computed","paper":{"title":"When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Liang Zhao, Xinyuan Song, Zekun Cai","submitted_at":"2026-06-26T07:35:43Z","abstract_excerpt":"Recursive self-training can degrade neural generative models when generated data is reused without fresh human data or external quality control. We study this risk in code LLMs, where AI-generated code can enter real repositories, later become training data, and create a repository-scale self-training loop. While software development traditionally interrupts this loop through pull-request review, tests, compilation, and human approval, AI coding tools now produce code faster than humans can review it, and code review itself is increasingly automated by AI systems. We therefore compare three re"},"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":"2606.28438","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-06-26T07:35:43Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7e9d0cf3bd1dc16f4ca202b2b675cf6b9fa6c5b3c475ea25c0870485cc5a081e","abstract_canon_sha256":"defcc7bf7b367c3318ec86df133b49e0d08ac35ca133ed71ac5109cf3f14dcff"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T00:15:14.054248Z","signature_b64":"xfBAQfqBNMixahcuURMr7yA6tSuv+s012clnxujI84TNipGdN3LUojCdDfnNVgcIHKVzGz6ht0mg1m0QMfGsAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0083e465e0853cffd238ed45f67d34c85fa7db335b9bc25cfa6859741cefe009","last_reissued_at":"2026-06-30T00:15:14.053871Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T00:15:14.053871Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Liang Zhao, Xinyuan Song, Zekun Cai","submitted_at":"2026-06-26T07:35:43Z","abstract_excerpt":"Recursive self-training can degrade neural generative models when generated data is reused without fresh human data or external quality control. We study this risk in code LLMs, where AI-generated code can enter real repositories, later become training data, and create a repository-scale self-training loop. While software development traditionally interrupts this loop through pull-request review, tests, compilation, and human approval, AI coding tools now produce code faster than humans can review it, and code review itself is increasingly automated by AI systems. We therefore compare three re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.28438","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/2606.28438/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":"2606.28438","created_at":"2026-06-30T00:15:14.053927+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.28438v1","created_at":"2026-06-30T00:15:14.053927+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.28438","created_at":"2026-06-30T00:15:14.053927+00:00"},{"alias_kind":"pith_short_12","alias_value":"ACB6IZPAQU6P","created_at":"2026-06-30T00:15:14.053927+00:00"},{"alias_kind":"pith_short_16","alias_value":"ACB6IZPAQU6P7URY","created_at":"2026-06-30T00:15:14.053927+00:00"},{"alias_kind":"pith_short_8","alias_value":"ACB6IZPA","created_at":"2026-06-30T00:15:14.053927+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/ACB6IZPAQU6P7URY5VC7M7JUZB","json":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB.json","graph_json":"https://pith.science/api/pith-number/ACB6IZPAQU6P7URY5VC7M7JUZB/graph.json","events_json":"https://pith.science/api/pith-number/ACB6IZPAQU6P7URY5VC7M7JUZB/events.json","paper":"https://pith.science/paper/ACB6IZPA"},"agent_actions":{"view_html":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB","download_json":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB.json","view_paper":"https://pith.science/paper/ACB6IZPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.28438&json=true","fetch_graph":"https://pith.science/api/pith-number/ACB6IZPAQU6P7URY5VC7M7JUZB/graph.json","fetch_events":"https://pith.science/api/pith-number/ACB6IZPAQU6P7URY5VC7M7JUZB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB/action/storage_attestation","attest_author":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB/action/author_attestation","sign_citation":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB/action/citation_signature","submit_replication":"https://pith.science/pith/ACB6IZPAQU6P7URY5VC7M7JUZB/action/replication_record"}},"created_at":"2026-06-30T00:15:14.053927+00:00","updated_at":"2026-06-30T00:15:14.053927+00:00"}