[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-new-way-to-catch-ai-blind-spots-using-counterfactuals":10,"sections":40},{"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":30,"tags":31,"sources":35,"feedback":39,"feedback_at":22,"cost_usd":39,"total_tokens":39},9259,"a-new-way-to-catch-ai-blind-spots-using-counterfactuals","A New Way to Catch AI Blind Spots Using Counterfactuals","A new technique uses counterfactual explanations to measure how far an input sits from a decision boundary, flagging data a model has never seen before.","A new out-of-distribution detector uses counterfactual explanations to measure how confused a model really is.\n\nResearchers built a post-hoc tool that spots when an AI system is looking at data unlike anything in its training set, a failure mode known as out-of-distribution, or OOD, input. Instead of just checking a confidence score, the method generates a counterfactual explanation - a minimal tweak to the input that would flip the model's decision - and measures how far away that tweak has to go. The farther the required change, the more likely the input is unfamiliar. Because computing those explanations is costly for large models, the team also worked out how to run the calculation directly in a model's internal embedding space, cutting the overhead.\n\nThe results give a concrete yardstick: on CIFAR-100 the method hit 97.05% AUROC (area under the ROC curve, a measure of how well the detector separates known data from unknown data, where 100% is perfect) and 13.79% FPR95 (the rate of false alarms when the detector is tuned to catch 95% of true unknowns, where lower is better). It matched existing methods on CIFAR-10 at 93.50% AUROC and beat them on the harder ImageNet-200 benchmark at 92.55% AUROC. That is a meaningful gap, since most detectors get worse, not better, as the number of classes grows.\n\nIt is not a universal fix for AI's blind spots, but it is one of the few OOD methods that also hands you a human-readable reason for its verdict, instead of just a suspicious-looking number.","[\"ai\",\"machine-learning\",\"model-safety\",\"research\"]","2026-10-01T04:00:00.000Z","2026-10-02T05:17:55.385Z","2026-10-02T05:17:58.286Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Define AUROC and FPR95 on first use (e.g., AUROC = area under the ROC curve measuring detection accuracy, FPR95 = false-positive rate at 95% true-positive rate) since the article cites these metrics as key results without ever explaining what they mean.","resolved","ai",[30,32,33,34],"machine-learning","model-safety","research",[36],{"name":37,"url":38},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.10148",0,{"sections":41},[42,45,49,53,58,63,67,72,77,81,86,91,96,101],{"name":43,"slug":30,"count":44,"latest_published_at":18},"AI",5659,{"name":46,"slug":47,"count":48,"latest_published_at":18},"Security","security",818,{"name":50,"slug":51,"count":52,"latest_published_at":18},"Policy","policy",430,{"name":54,"slug":55,"count":56,"latest_published_at":57},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":62},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":64,"slug":65,"count":66,"latest_published_at":18},"Science","science",163,{"name":68,"slug":69,"count":70,"latest_published_at":71},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":78,"slug":79,"count":75,"latest_published_at":80},"Software","software","2026-09-30T21:41:11.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":102,"slug":103,"count":104,"latest_published_at":105},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]