Step 6: tool results

read_page

{
  "url": "https://arxiv.org/abs/2512.07810",
  "max_chars": 2600
}

{"ok":true,"url":"https://arxiv.org/abs/2512.07810","text":"[2512.07810] Auditing Games for Sandbagging\nSkip to main content\nSearch\nSubmit\nDonate\nLog in\nSearch arXiv\nPress Enter to search · Advanced search\n-->\nComputer Science > Artificial Intelligence\narXiv:2512.07810 (cs)\n[Submitted on 8 Dec 2025]\nTitle: Auditing Games for Sandbagging\nAuthors: Jordan Taylor , Sid Black , Dillon Bowen , Thomas Read , Satvik Golechha , Alex Zelenka-Martin , Oliver Makins , Connor Kissane , Kola Ayonrinde , Jacob Merizian , Samuel Marks , Chris Cundy , Joseph Bloom\nView a PDF of the paper titled Auditing Games for Sandbagging, by Jordan Taylor and 12 other authors\nView PDF\nHTML (experimental)\nAbstract: Future AI systems could conceal their capabilities ('sandbagging') during evaluations, potentially misleading developers and auditors. We stress-tested sandbagging detection techniques using an auditing game. First, a red team fine-tuned five models, some of which conditionally underperformed, as a proxy for sandbagging. Second, a blue team used black-box, model-internals, or training-based approaches to identify sandbagging models. We found that the blue team could not reliably discriminate sandbaggers from benign models. Black-box approaches were defeated by effective imitation of a weaker model. Linear probes, a model-internals approach, showed more promise but their naive application was vulnerable to behaviours instilled by the red team. We also explored capability elicitation as a strategy for detecting sandbagging. Although Prompt-based elicitation was not reliable, training-based elicitation consistently elicited full performance from the sandbagging models, using only a single correct demonstration of the evaluation task. However the performance of benign models was sometimes also raised, so relying on elicitation as a detection strategy was prone to false-positives. In the short-term, we recommend developers remove potential sandbagging using on-distribution training for elicitation. In the longer-term, further research is needed to ensure the efficacy of training-based elicitation, and develop robust methods for sandbagging detection. We open source our model organisms at this https URL and select transcripts and results at this https URL . A demo illustrating the game can be played at this https URL .\nComments:\n77 pages (28 non-appendix pages), 38 figures\nSubjects:\nArtificial Intelligence (cs.AI)\nCite as:\narXiv:2512.07810 [cs.AI]\n(or\narXiv:2512.07810v1 [cs.AI] for this version)\nhttps://doi.org/10.48550/arXiv.2512.07810\nFocus to learn more\narXiv-issued DOI via DataCite\nSubmission history\nFrom: Jordan Taylor [ view ","chars":2600,"total_chars":5094,"offset":0,"truncated":true,"next_offset":2600}