[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-teaching-llms-to-build-shorthand-like-humans-do":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},2872,"teaching-llms-to-build-shorthand-like-humans-do","Teaching LLMs to Build Shorthand Like Humans Do","A new post-training method helps language models develop ad-hoc communication conventions over multi-turn conversations, closing a gap with human behavior.","Language models can now be fine-tuned to do something humans do naturally: develop shared shorthand over the course of a conversation.\n\nResearchers published a post-training process that targets convention formation — the way people streamline language across repeated interactions, settling on abbreviations, nicknames, and implied references that speed things up. Prior work had established that LLMs don't naturally converge on these conventions the way humans do. The new approach fine-tunes models on examples of convention formation identified through heuristics, rather than hand-labeling. The team also built two benchmarks to measure the ability: one cognitively-motivated interaction task designed to reliably elicit convention formation in humans, and a document-grounded reference completion task meant to reflect real-world usage.\n\nThe gap this addresses is more practical than it sounds. Most LLM deployments assume each conversation is stateless, but long-running agent workflows and multi-session tools increasingly depend on models that can compress context efficiently. A model that never builds shared shorthand with a user — or with another model in a multi-agent pipeline — wastes tokens and forces repetition.\n\nThe benchmark work here is at least as interesting as the training method itself. Convention formation has been hard to measure precisely; building two distinct evaluation frameworks gives the field something to check future claims against. Whether this transfers cleanly outside lab conditions, or holds up when the \"convention\" involves ambiguous or domain-specific terminology, is the next obvious question.","[\"ai\",\"language-models\",\"research\",\"post-training\"]","2026-06-30T04:00:00.000Z","2026-06-30T14:21:39.298Z","2026-06-30T14:21:42.212Z","published",null,[],"ai",[24,26,27,28],"language-models","research","post-training",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.06482",0,{"sections":35},[36,40,45,50,55,60,65,70,75,80,85,89,94,99],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":86,"slug":87,"count":83,"latest_published_at":88},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",29,"2026-07-10T22:28:58.000Z",{"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"]