{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VHSAKT2JCX2O5SP2KJGCCQIWAO","short_pith_number":"pith:VHSAKT2J","schema_version":"1.0","canonical_sha256":"a9e4054f4915f4eec9fa524c21411603841b0d266d3c2bd0e81edad8144dec7c","source":{"kind":"arxiv","id":"2608.09643","version":1},"attestation_state":"computed","paper":{"title":"Activation Probes Surface Code-Security Signals that the Model's Output Misses","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Ivan Wiryadi","submitted_at":"2026-08-10T14:22:46Z","abstract_excerpt":"AI coding agents now write a growing share of production code, and human security review does not scale at the rate code is generated. The agents in widest use are closed-weight, so a deploying team cannot read their internals. It can instead run an open-weight model as a reviewer over the agent's output. That reviewer's activations are readable. We ask whether reading those activations recovers a security signal that simply asking the same reviewer misses. We fit a single linear probe per model on a corpus of paired vulnerable-and-fixed Python functions, then test it without retraining on rea"},"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":"2608.09643","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-08-10T14:22:46Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f480addc611cb5f5a5296968a254d11fba47e659adace184cbb8fb3f7f96b2e0","abstract_canon_sha256":"f434690f4964c50dc57ec4057d2ea0545eac738d4b2a20d237e3fa08662fb78f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T02:24:43.896520Z","signature_b64":"DA5H88/WBDBMlvqBO33rs8NBwzlpsa60GFDfHeZ2LxQBfSP/dMDWAwLbX7PnxRaAzPZ+nrR6L43fJmRSa36ICA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a9e4054f4915f4eec9fa524c21411603841b0d266d3c2bd0e81edad8144dec7c","last_reissued_at":"2026-08-11T02:24:43.894922Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T02:24:43.894922Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Activation Probes Surface Code-Security Signals that the Model's Output Misses","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Ivan Wiryadi","submitted_at":"2026-08-10T14:22:46Z","abstract_excerpt":"AI coding agents now write a growing share of production code, and human security review does not scale at the rate code is generated. The agents in widest use are closed-weight, so a deploying team cannot read their internals. It can instead run an open-weight model as a reviewer over the agent's output. That reviewer's activations are readable. We ask whether reading those activations recovers a security signal that simply asking the same reviewer misses. We fit a single linear probe per model on a corpus of paired vulnerable-and-fixed Python functions, then test it without retraining on rea"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.09643","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/2608.09643/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":"2608.09643","created_at":"2026-08-11T02:24:43.895554+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.09643v1","created_at":"2026-08-11T02:24:43.895554+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.09643","created_at":"2026-08-11T02:24:43.895554+00:00"},{"alias_kind":"pith_short_12","alias_value":"VHSAKT2JCX2O","created_at":"2026-08-11T02:24:43.895554+00:00"},{"alias_kind":"pith_short_16","alias_value":"VHSAKT2JCX2O5SP2","created_at":"2026-08-11T02:24:43.895554+00:00"},{"alias_kind":"pith_short_8","alias_value":"VHSAKT2J","created_at":"2026-08-11T02:24:43.895554+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/VHSAKT2JCX2O5SP2KJGCCQIWAO","json":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO.json","graph_json":"https://pith.science/api/pith-number/VHSAKT2JCX2O5SP2KJGCCQIWAO/graph.json","events_json":"https://pith.science/api/pith-number/VHSAKT2JCX2O5SP2KJGCCQIWAO/events.json","paper":"https://pith.science/paper/VHSAKT2J"},"agent_actions":{"view_html":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO","download_json":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO.json","view_paper":"https://pith.science/paper/VHSAKT2J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.09643&json=true","fetch_graph":"https://pith.science/api/pith-number/VHSAKT2JCX2O5SP2KJGCCQIWAO/graph.json","fetch_events":"https://pith.science/api/pith-number/VHSAKT2JCX2O5SP2KJGCCQIWAO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO/action/storage_attestation","attest_author":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO/action/author_attestation","sign_citation":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO/action/citation_signature","submit_replication":"https://pith.science/pith/VHSAKT2JCX2O5SP2KJGCCQIWAO/action/replication_record"}},"created_at":"2026-08-11T02:24:43.895554+00:00","updated_at":"2026-08-11T02:24:43.895554+00:00"}