[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-mbr-decoding-gets-a-noisy-channel-upgrade":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},4019,"mbr-decoding-gets-a-noisy-channel-upgrade","MBR Decoding Gets a Noisy-Channel Upgrade","Researchers decompose minimum Bayes risk decoding into four components, potentially improving how AI models select their best output.","A new paper proposes reframing how language models pick their final output by injecting a noisy-channel decomposition into minimum Bayes risk decoding.\n\nMinimum Bayes risk (MBR) decoding is an alternative to the standard maximum a posteriori (MAP) approach. Instead of picking the single most probable output, MBR samples a set of candidate outputs and selects the one that scores best against the others on a utility metric like BLEU or COMET. The problem the authors identify: those metrics are asymmetric — they score differently depending on which text is treated as the hypothesis and which as the reference. Standard MBR ignores that directionality. The new framework splits the decoding process into four components: hypothesis-to-reference likelihood, reference-to-hypothesis likelihood, hypothesis prior, and reference prior.\n\nThat four-way decomposition matters because it makes explicit what existing MBR variants were already doing implicitly, and it gives researchers a lever to tune. The paper reports that channel-wise contributions behave differently across metrics but stay consistent across tasks — meaning the right weighting could squeeze out better performance without changing the underlying model at all.\n\nMBR decoding has been gaining traction as inference budgets grow and teams look for ways to improve output quality without retraining. This work does not replace MBR; it offers a cleaner theoretical scaffold for the variants that already exist and a principled reason to expect that tuning channel weights will help. Whether it translates to meaningful gains on real tasks remains to be seen in follow-on work.","[\"ai\",\"nlp\",\"decoding\",\"research\"]","2026-07-07T04:00:00.000Z","2026-07-07T15:08:07.970Z","2026-07-07T15:08:10.798Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"The headline and dek are vague and placeholder-like — 'A Better Way to Pick What an AI Actually Says' does not name MBR decoding, noisy-channel decomposition, or any concrete claim, failing the requirement that the headline clearly states the actual news.","resolved","ai",[30,32,33,34],"nlp","decoding","research",[36],{"name":37,"url":38},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.05198",0,{"sections":41},[42,46,51,56,61,66,71,76,81,85,90,94,99,104],{"name":43,"slug":30,"count":44,"latest_published_at":45},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":57,"slug":58,"count":59,"latest_published_at":60},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":62,"slug":63,"count":64,"latest_published_at":65},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":75},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":18},"Dev Tools","dev-tools",59,{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":91,"slug":92,"count":88,"latest_published_at":93},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":105,"slug":106,"count":107,"latest_published_at":108},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]