[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-better-way-to-prune-mixture-of-experts-ai-models":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},6644,"a-better-way-to-prune-mixture-of-experts-ai-models","A Better Way to Prune Mixture-of-Experts AI Models","A new pruning technique called HOPE accounts for how AI model experts work together, beating prior methods especially at aggressive compression rates.","Shrinking a giant AI model without dumbing it down just got a bit easier.\n\nMixture-of-experts (MoE) language models split their work across many specialized sub-networks, called experts, and activate only a few per token. That trick still leaves models with huge total parameter counts to store in memory, so researchers prune away lower-value experts to cut costs. Most pruning methods score each expert on its own and assume their contributions simply add up. A new paper introduces HOPE (Higher-Order Pruning of Experts), which instead models how experts work together, using a second-order objective that accounts for interactions between them. Tested on three frontier MoE models as large as 122 billion parameters, HOPE beat REAP, the previous best pruning method, especially at aggressive 50% pruning rates, scoring an average rank of 1.58 versus REAP's 2.42 across five methods, with gains of up to 6.1% on agentic coding benchmarks.\n\nWhy it matters: memory, not compute, is often the real bottleneck for running large MoE models, and pruning is one of the few levers that shrinks that footprint directly. HOPE's gains grow precisely where it counts most, on complex, multi-step tasks like agentic coding, where models lean on varied combinations of experts rather than the same ones repeatedly.\n\nIt's a solid, well-benchmarked improvement, not a breakthrough. HOPE is explicitly built as a generalization of REAP, and the paper frames existing first-order pruning as a special case it subsumes: progress by refinement, not reinvention.","[\"mixture-of-experts\",\"model-compression\",\"ai-research\",\"efficiency\"]","2026-09-17T04:00:00.000Z","2026-09-18T04:06:53.277Z","2026-09-18T04:07:05.204Z","published",null,[],"ai",[26,27,28,29],"mixture-of-experts","model-compression","ai-research","efficiency",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.18916",0,{"sections":36},[37,41,45,50,55,59,63,68,73,77,82,87,92,97],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",3853,"2026-09-17T08:27:09.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":18},"Security","security",648,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",338,"2026-09-11T04:00:00.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":18},"Hardware","hardware",154,{"name":60,"slug":61,"count":62,"latest_published_at":18},"Science","science",114,{"name":64,"slug":65,"count":66,"latest_published_at":67},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":69,"slug":70,"count":71,"latest_published_at":72},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":74,"slug":75,"count":76,"latest_published_at":18},"Dev Tools","dev-tools",73,{"name":78,"slug":79,"count":80,"latest_published_at":81},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"General","general",41,"2026-09-08T01:57:23.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]