Step 6: tool results

{
  "url": "https://icml.cc/virtual/2026/poster/61923",
  "title": "Same Question, Different Lies: Cross-Context Consistency (C³) for Black-Box Sandbagging Detection (ICML 2026 poster, lay summary)",
  "stance": "affirms",
  "context": "Lay summary of the paper introducing C³; the abstract itself says more cautiously that C³ maintains 'a classification signal where other black-box methods fail' across prompted and fine-tuned scenarios.",
  "speaker": "Yulong Lin, Pablo Bernabeu-Pérez, Benjamin Arnav, Lennie Wells, Mary Phuong",
  "claim_id": "37c8f4aa-8784-403b-979f-fc06db79597c",
  "confidence": 0.6,
  "publication": "ICML 2026",
  "source_date": "2026-07",
  "source_kind": "conference_paper",
  "verbatim_text": "We tested C³ across a range of sandbagging scenarios and showed that it reliably detected fake incompetence even where existing methods failed, offering a practical tool to help ensure AI safety evaluations can be trusted."
}

{
  "url": "https://arxiv.org/html/2611.14950",
  "title": "Stress-Testing Sandbagging Detection Under Adaptive Attack",
  "stance": "mentions",
  "context": "Related work; the paper shows a text-only sandbagging audit falling from AUC 0.978 to 0.560 under two rounds of adaptive retraining, and notes C³ has not yet faced such an attack.",
  "speaker": "Bhoj Rani Soopal, Aarohi Agrawal, Josh Karthikeyan, Edward Lue Chee Lip",
  "claim_id": "37c8f4aa-8784-403b-979f-fc06db79597c",
  "source_date": "2026-11-11",
  "source_kind": "preprint",
  "verbatim_text": "C 3 , which looks for inconsistent answers across paraphrases ( Lin et al., 2026 ) . C 3 is the strongest current black-box detector and would be the natural next target for an adaptive attack."
}

{
  "url": "https://lacuna.tiptreesystems.com/work/when-lie-detectors-learn-model-identity-confounds-in-black-box-sandbagging/wrk_cc46e467031489731dcb56615007b479",
  "title": "When Lie Detectors Learn Model Identity: Confounds in Black-Box Sandbagging Detection (summary)",
  "stance": "denies",
  "context": "Summary of the workshop companion to the C³ paper; reports moderate recall for fine-tuned (password-locked) sandbaggers, which cuts against high recall across fine-tuned cases.",
  "claim_id": "37c8f4aa-8784-403b-979f-fc06db79597c",
  "confidence": 0.6,
  "publication": "Lacuna",
  "source_kind": "webpage",
  "verbatim_text": "Password-locked sandbaggers: C³ achieved 52.9% recall on WMDP and 36.9% on GSM8K, outperforming other methods but still facing challenges."
}