[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-cheaper-way-to-fine-tune-llms-with-evolution-strategies":10,"sections":34},{"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":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},8160,"a-cheaper-way-to-fine-tune-llms-with-evolution-strategies","A Cheaper Way to Fine-Tune LLMs With Evolution Strategies","A new analysis of the EGGROLL training trick reveals hidden instability risks, and a proposed fix boosts math benchmark scores by up to 14 points.","Evolution strategies just got a tune-up, and the fix comes from finding a flaw in the shortcut that made them fast enough for large language models.\n\nEarlier this year, a technique called EGGROLL made it feasible to train huge models with evolution strategies by using low-rank, often rank-one, random perturbations instead of full dense noise. That shortcut is what makes ES scale, but the new paper shows it also warps the underlying math: the rank-one trick can turn the algorithm's effective update into something that no longer behaves like a clean gradient, occasionally flipping the stability of an optimum it should be converging toward. The researchers also found the extra noise from using rank-one instead of dense perturbations shrinks fast as models get wider, down to just 0.098% at a width of 4096. From that diagnosis they built LOO-ROLL, an estimator that gets the same signal from one evaluation per direction instead of two.\n\nThat efficiency gain is not just theoretical. Tested across fourteen post-training runs up to 14 billion parameters, LOO-ROLL beat EGGROLL on eleven paired comparisons with no losses, and on math benchmarks the gains were substantial: up to 14.1 points on GSM8K and 12.2 points on MATH-500 at matched compute time.\n\nIt is a reminder that the tricks making today's training methods fast are not free lunches, they are trading exactness for speed in ways that only get scrutinized after the fact.","[\"evolution-strategies\",\"llm-training\",\"ai-research\"]","2026-09-28T04:00:00.000Z","2026-09-28T11:55:15.756Z","2026-09-28T11:55:22.113Z","published",null,[],"ai",[26,27,28],"evolution-strategies","llm-training","ai-research",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.10980",0,{"sections":35},[36,39,43,48,53,58,62,67,72,77,82,87,91,96],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",4799,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",762,{"name":44,"slug":45,"count":46,"latest_published_at":47},"Policy","policy",399,"2026-09-27T18:39:02.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",261,"2026-09-27T15:30:35.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Hardware","hardware",188,"2026-09-27T20:46:36.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",151,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",135,"2026-09-26T14:30:00.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Dev Tools","dev-tools",84,"2026-09-26T04:20:58.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":88,"slug":89,"count":85,"latest_published_at":90},"General","general","2026-09-26T17:02:42.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]