[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-finetuned-ai-models-contradict-themselves-when-resampled":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},10913,"finetuned-ai-models-contradict-themselves-when-resampled","Finetuned AI Models Contradict Themselves When Resampled","A new benchmark shows narrowly finetuned AI models contradict themselves on repeat answers, complicating claims about measuring misalignment.","A new arXiv paper finds that narrowly finetuned AI models contradict themselves when you ask them the same question twice.\n\nResearchers built a set of 175 questions specifically designed so that different answers across repeated samples cannot be explained away by ambiguity in the question or genuine indifference between options. They used it to measure incoherence: how often a model flatly contradicts its own previous answer when resampled. The method is deliberately conservative, flagging a model as incoherent only when the contradiction is glaring rather than just noisy. Even with that high bar, models built through narrow finetuning, the kind of specialized test subjects researchers use to study misaligned behavior, still failed often, showing identity confusion, failed introspection, and after-the-fact rationalizations for answers that did not add up.\n\nThat is a problem because a chunk of AI safety research leans on these narrow finetunes as stand-ins for studying how misalignment shows up and behaves. If the underlying models are this inconsistent, it gets hard to tell whether the misaligned patterns researchers report are a real signal or just noise from a model that cannot keep its story straight. The paper does not claim fraud or bad faith, just that the test subjects may be too confused to trust.\n\nIt is the research equivalent of discovering your lab mice do not reliably remember their own names, worth knowing before drawing big conclusions from their behavior.","[\"ai safety\",\"model alignment\",\"ai research\",\"llm evaluation\"]","2026-10-09T04:00:00.000Z","2026-10-09T21:16:18.081Z","2026-10-09T21:16:24.333Z","published",null,[],"ai",[26,27,28,29],"ai safety","model alignment","ai research","llm evaluation",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12129",0,{"sections":36},[37,40,44,49,54,58,62,67,72,77,82,87,92,97],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",6690,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",930,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":18},"Hardware","hardware",231,{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",192,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]