{"paper":{"title":"AtomEval: Validity-Aware Atomic Evaluation of Adversarial Claim Rewriting in Fact Verification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"AtomEval evaluates adversarial claims by decomposing them into atomic SROM components to better detect factual corruptions.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Hanze Jia, Hongyi Cen, Jingyi Zheng, Mingxin Wang, Tan Tang, Yule Liu","submitted_at":"2026-04-09T08:32:35Z","abstract_excerpt":"Large language models (LLMs) can rewrite refuted claims to evade evidence-based fact verifiers, but conventional attack success rate (ASR) can be inflated when rewrites change, weaken, or correct the false proposition they are supposed to preserve. We introduce AtomEval, a validity-aware evaluation protocol for fixed-evidence adversarial claim rewriting. AtomEval represents claims as subject--relation--object--modifier (SROM) atoms, applies a one-way preservation gate to separate valid verifier evasion from proposition-changing rewrites, and reports validity-aware attack success rate (VASR), w"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on the FEVER dataset across representative attack strategies and LLM generators show that AtomEval provides more reliable evaluation signals in our experiments.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That decomposing claims into SROM atoms and applying Atomic Validity Scoring accurately captures truth-conditional consistency and detects factual corruption without introducing its own biases or missing key semantic nuances.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AtomEval introduces atomic claim decomposition and validity scoring to provide more reliable evaluation of adversarial rewrites than standard similarity metrics in fact verification.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AtomEval evaluates adversarial claims by decomposing them into atomic SROM components to better detect factual corruptions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"59d56a0cf3155d5850aa7c6166f060f02ad0866ef7057b78b97965f6a52a43cc"},"source":{"id":"2604.07967","kind":"arxiv","version":3},"verdict":{"id":"b032e29c-aa07-4a8f-b1b9-c78af1184fe4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T18:12:53.655768Z","strongest_claim":"Experiments on the FEVER dataset across representative attack strategies and LLM generators show that AtomEval provides more reliable evaluation signals in our experiments.","one_line_summary":"AtomEval introduces atomic claim decomposition and validity scoring to provide more reliable evaluation of adversarial rewrites than standard similarity metrics in fact verification.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That decomposing claims into SROM atoms and applying Atomic Validity Scoring accurately captures truth-conditional consistency and detects factual corruption without introducing its own biases or missing key semantic nuances.","pith_extraction_headline":"AtomEval evaluates adversarial claims by decomposing them into atomic SROM components to better detect factual corruptions."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.07967/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":2,"snapshot_sha256":"e0832def088f423003060e63522f38e83daf1b5784d2a3a96111b76d11d49c7b"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}