[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-benchmark-tests-how-well-ai-handles-bioengineering-problems":10,"sections":41},{"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":36,"feedback":40,"feedback_at":22,"cost_usd":40,"total_tokens":40},8180,"new-benchmark-tests-how-well-ai-handles-bioengineering-problems","New Benchmark Tests How Well AI Handles Bioengineering Problems","A 22-institution benchmark found top AI models hit 90% accuracy on bioengineering questions but struggled more on literature synthesis and image-based tasks.","Researchers from 22 institutions built a 608-item exam to see if AI models can actually reason through bioengineering problems, not just recall facts.\n\nBioEVAL was assembled by those 22 research groups across 11 bioengineering subfields, plus a catch-all category for items that didn't fit neatly anywhere. The benchmark started with 608 items: 380 multiple-choice questions, 218 literature-synthesis tasks, and 10 multimodal problems requiring interpretation of experimental lab images. After evaluation, a blinded cross-group audit of the highest- and lowest-scoring MCQs flagged 21 questions for revision or removal, so all reported MCQ scores are based on the 359 questions that survived that cut. The team then tested cloud models including ChatGPT, Gemini, and Grok alongside smaller models built to run on consumer-grade GPUs.\n\nTop models hit 90% accuracy on the retained MCQs, but sagged to a 0.72 similarity score on literature synthesis and 80% on the small multimodal sample. That spread matters because multiple-choice recall is the easy part - actual bioengineering work involves reconciling conflicting papers and reading experimental images, exactly where scores dropped.\n\nThe audit step - which quietly dropped 21 of the original 380 multiple-choice questions after the fact - is a useful admission that even a PhD-level benchmark needs its own quality control before anyone trusts the leaderboard.","[\"bioengineering\",\"ai-benchmarks\",\"llm-evaluation\",\"multimodal-ai\"]","2026-09-28T04:00:00.000Z","2026-09-28T13:00:14.143Z","2026-09-28T13:00:20.754Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Fix the arithmetic mismatch: the dek\u002Fbody say '608 items' but list 359 MCQs + 218 literature synthesis + 10 multimodal = 587 — the source's 608 total is 380 pre-audit MCQs + 218 + 10, with 359 being the post-audit retained count, so clarify that distinction instead of conflating the two totals.","resolved","ai",[32,33,34,35],"bioengineering","ai-benchmarks","llm-evaluation","multimodal-ai",[37],{"name":38,"url":39},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.30489",0,{"sections":42},[43,46,50,55,60,65,69,74,79,84,89,94,98,103],{"name":44,"slug":30,"count":45,"latest_published_at":18},"AI",4798,{"name":47,"slug":48,"count":49,"latest_published_at":18},"Security","security",762,{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",399,"2026-09-27T18:39:02.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Deals","deals",261,"2026-09-27T15:30:35.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Hardware","hardware",188,"2026-09-27T20:46:36.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":18},"Science","science",151,{"name":70,"slug":71,"count":72,"latest_published_at":73},"Consumer Tech","consumer-tech",135,"2026-09-26T14:30:00.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Dev Tools","dev-tools",84,"2026-09-26T04:20:58.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":95,"slug":96,"count":92,"latest_published_at":97},"General","general","2026-09-26T17:02:42.000Z",{"name":99,"slug":100,"count":101,"latest_published_at":102},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":104,"slug":105,"count":106,"latest_published_at":107},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]