{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:I6VRWGKDI7I6N6P32R5CYSANXT","short_pith_number":"pith:I6VRWGKD","schema_version":"1.0","canonical_sha256":"47ab1b194347d1e6f9fbd47a2c480dbccc8d58b71af548b271cf66d22f27e14b","source":{"kind":"arxiv","id":"2605.17450","version":1},"attestation_state":"computed","paper":{"title":"ContraFix: Agentic Vulnerability Repair via Differential Runtime Evidence and Skill Reuse","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ContraFix identifies root causes for vulnerabilities by comparing state differences in crashing versus non-crashing PoC variants and reuses prior repair skills.","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.SE","authors_text":"Fang Liu, Li Zhang, Simiao Liu, Yang Liu, Yinghao Zhu","submitted_at":"2026-05-17T13:48:25Z","abstract_excerpt":"Large language model (LLM) agents are increasingly used for automated vulnerability repair (AVR), where repository-level reasoning enables them to inspect context and produce source-code patches. However, recent empirical results show that these agents still struggle with real-world vulnerabilities. Their main failure mode is semantic misunderstanding: choosing a repair direction that does not match the root cause. We identify two reasons for this gap. Existing agents usually reason from the failing execution alone. A crash report can pinpoint where the program failed, but it does not reveal w"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.17450","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2026-05-17T13:48:25Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR"],"title_canon_sha256":"b9ea41076ae752434094f0daf0db13afb0b6f94f04b53e6d879b07296c4078aa","abstract_canon_sha256":"8e2d18aa448da2a49ec500383f1f7344616231d5785e40f1e5155d004a4761d0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:39.605852Z","signature_b64":"AOOBo27VszJzX1isV+15uHdZMMhmLAk1n+3dtJewR4GMTPlu2O8uP/AVC2nkdbGHktGTwyugL9sNjxuK1KLPDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"47ab1b194347d1e6f9fbd47a2c480dbccc8d58b71af548b271cf66d22f27e14b","last_reissued_at":"2026-05-20T00:04:39.605167Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:39.605167Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ContraFix: Agentic Vulnerability Repair via Differential Runtime Evidence and Skill Reuse","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ContraFix identifies root causes for vulnerabilities by comparing state differences in crashing versus non-crashing PoC variants and reuses prior repair skills.","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.SE","authors_text":"Fang Liu, Li Zhang, Simiao Liu, Yang Liu, Yinghao Zhu","submitted_at":"2026-05-17T13:48:25Z","abstract_excerpt":"Large language model (LLM) agents are increasingly used for automated vulnerability repair (AVR), where repository-level reasoning enables them to inspect context and produce source-code patches. However, recent empirical results show that these agents still struggle with real-world vulnerabilities. Their main failure mode is semantic misunderstanding: choosing a repair direction that does not match the root cause. We identify two reasons for this gap. Existing agents usually reason from the failing execution alone. A crash report can pinpoint where the program failed, but it does not reveal w"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On SEC-Bench (C/C++, 200 instances) and PatchEval (Go, Python, JavaScript, 225 instances), ContraFix with GPT-5-mini resolves 84.0% and 73.8% of the tasks, respectively, achieving state-of-the-art performance on both benchmarks while costing less than one-third of the strongest comparable baseline.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the divergences identified by state probes between crashing and non-crashing PoC variants reliably isolate the causal variables or state transitions responsible for the vulnerability, and that these can be converted into a repair specification that produces correct, verified patches rather than symptom fixes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ContraFix couples differential runtime evidence from execution variants with reusable repair skills to achieve 84.0% resolution on SEC-Bench and 73.8% on PatchEval using GPT-5-mini, outperforming baselines at lower cost.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ContraFix identifies root causes for vulnerabilities by comparing state differences in crashing versus non-crashing PoC variants and reuses prior repair skills.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"55fca1f5fb0d34845a6352af565aeacfc9b8dc8fe27785e074b8f4165c591e48"},"source":{"id":"2605.17450","kind":"arxiv","version":1},"verdict":{"id":"c850268c-533d-41ee-961c-828078fa3efc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:57:32.906485Z","strongest_claim":"On SEC-Bench (C/C++, 200 instances) and PatchEval (Go, Python, JavaScript, 225 instances), ContraFix with GPT-5-mini resolves 84.0% and 73.8% of the tasks, respectively, achieving state-of-the-art performance on both benchmarks while costing less than one-third of the strongest comparable baseline.","one_line_summary":"ContraFix couples differential runtime evidence from execution variants with reusable repair skills to achieve 84.0% resolution on SEC-Bench and 73.8% on PatchEval using GPT-5-mini, outperforming baselines at lower cost.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the divergences identified by state probes between crashing and non-crashing PoC variants reliably isolate the causal variables or state transitions responsible for the vulnerability, and that these can be converted into a repair specification that produces correct, verified patches rather than symptom fixes.","pith_extraction_headline":"ContraFix identifies root causes for vulnerabilities by comparing state differences in crashing versus non-crashing PoC variants and reuses prior repair skills."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17450/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T23:31:19.945467Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T23:12:24.886442Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:41:57.714109Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.666915Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"8abf52f601c599bc470f0931f46af43b678c75da95416b936437151bfd7a30c0"},"references":{"count":61,"sample":[{"doi":"","year":2025,"title":"SEC-bench/aider","work_id":"7b424c31-93c0-4afa-ab2a-9d0486abd61e","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Amir Al-Maamari. 2026. Why LLMs Fail: A Failure Analysis and Partial Success Measurement for Automated Security Patch Generation. arXiv:2603.10072 [cs.CR] https://arxiv.org/abs/2603.10072","work_id":"5b08e7d9-f7e1-4487-ba8a-f37a81884879","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Anthropic. 2025. Claude Code: A Command Line Tool for Agentic Coding. https://code.claude.com/docs. Accessed: 2026-03-26","work_id":"1a47890a-24ad-42cd-8a69-b07ca108a166","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Afsah Anwar, Aminollah Khormali, Hisham Alasmary, Sung J Choi, Saeed Salem, David Mohaisen, et al. 2020. Measuring the Cost of Software Vulnerabilities.EAI Endorsed Transactions on Security & Safety7,","work_id":"d88d9ad8-f85b-46f8-93dd-04603dcafaa2","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Tim Blazytko, Moritz Schlögel, Cornelius Aschermann, Ali Abbasi, Joel Frank, Simon Wörner, and Thorsten Holz. 2020. AURORA: statistical crash analysis for automated root cause explanation. InProceedin","work_id":"e04e3d0d-4c8b-407c-aac4-b062fbeb8b84","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":61,"snapshot_sha256":"6d5a92d3a59506f2ac2b072ea741dd92e6394dfeba1254a8f4e50b2aa0aa2cd3","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f177f11f332fb9c85c58a78c81e46fc408eb93d1ad53df0387505126e2b75081"},"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":"2605.17450","created_at":"2026-05-20T00:04:39.605299+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17450v1","created_at":"2026-05-20T00:04:39.605299+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17450","created_at":"2026-05-20T00:04:39.605299+00:00"},{"alias_kind":"pith_short_12","alias_value":"I6VRWGKDI7I6","created_at":"2026-05-20T00:04:39.605299+00:00"},{"alias_kind":"pith_short_16","alias_value":"I6VRWGKDI7I6N6P3","created_at":"2026-05-20T00:04:39.605299+00:00"},{"alias_kind":"pith_short_8","alias_value":"I6VRWGKD","created_at":"2026-05-20T00:04:39.605299+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT","json":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT.json","graph_json":"https://pith.science/api/pith-number/I6VRWGKDI7I6N6P32R5CYSANXT/graph.json","events_json":"https://pith.science/api/pith-number/I6VRWGKDI7I6N6P32R5CYSANXT/events.json","paper":"https://pith.science/paper/I6VRWGKD"},"agent_actions":{"view_html":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT","download_json":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT.json","view_paper":"https://pith.science/paper/I6VRWGKD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17450&json=true","fetch_graph":"https://pith.science/api/pith-number/I6VRWGKDI7I6N6P32R5CYSANXT/graph.json","fetch_events":"https://pith.science/api/pith-number/I6VRWGKDI7I6N6P32R5CYSANXT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT/action/storage_attestation","attest_author":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT/action/author_attestation","sign_citation":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT/action/citation_signature","submit_replication":"https://pith.science/pith/I6VRWGKDI7I6N6P32R5CYSANXT/action/replication_record"}},"created_at":"2026-05-20T00:04:39.605299+00:00","updated_at":"2026-05-20T00:04:39.605299+00:00"}