[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-language-models-favor-svo-word-order-as-training-data-grows":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},5352,"language-models-favor-svo-word-order-as-training-data-grows","Language Models Favor SVO Word Order as Training Data Grows","A new study finds LLMs increasingly favor SVO word order as training data grows, a data-driven bias that could erode global word order diversity.","Large language models are quietly drifting toward English-style sentence order as they get bigger and better-fed, regardless of whether that order reflects how most human languages actually work.\n\nResearchers tested decoder-only language models on 192 artificial languages built to isolate word-order effects, plus a set of typologically diverse real ones. On the artificial languages, models leaned left-branching, a pattern that matches neither human language universals nor how people actually learn word order. On real languages, small models showed no consistent preference. But as training data scaled up, both monolingual and multilingual models increasingly favored subject-verb-object structure, the order English uses, even though subject-object-verb is the most common order cross-linguistically. The SVO tilt tracked with how much and how well-curated a language's training data was, not with any inherent property of the word order itself.\n\nThat's the real finding here: word-order bias in these models isn't architectural, it's a byproduct of whose text got scraped and cleaned. Because a small cluster of high-resource languages, nearly all SVO, dominates the data supply, the paper warns that widespread LLM adoption could gradually flatten word-order diversity worldwide, hitting hardest in languages that flexibly use multiple word orders.\n\nCall it algorithmic prestige bias. It's the same mechanism that made English-inflected phrasing creep into machine translation output for a decade, just showing up one abstraction level down, in the grammar itself rather than the vocabulary.","[\"language models\",\"linguistics\",\"nlp research\",\"word order\"]","2026-08-18T04:00:00.000Z","2026-08-18T16:34:03.260Z","2026-08-18T16:34:15.227Z","published",null,[],"ai",[26,27,28,29],"language models","linguistics","nlp research","word order",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.15129",0,{"sections":36},[37,41,45,50,55,60,65,70,75,79,84,89,94,99],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",3293,"2026-08-20T04:00:00.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":40},"Security","security",435,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",210,"2026-08-19T09:32:27.000Z",{"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":59},"Hardware","hardware",140,"2026-08-19T18:25:42.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",95,"2026-08-18T16:05:00.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Science","science",90,"2026-08-19T18:41:02.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",73,"2026-08-18T07:51:50.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":18},"Dev Tools","dev-tools",69,{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",47,"2026-08-19T19:13:46.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",33,"2026-08-18T22:18:13.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]