[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-llms-now-build-hardware-knowledge-graphs-from-datasheets":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},11258,"llms-now-build-hardware-knowledge-graphs-from-datasheets","LLMs Now Build Hardware Knowledge Graphs From Datasheets","A new LLM pipeline extracts hardware specs from datasheets into knowledge graphs, catching far more detail than a single one-shot prompt.","A new pipeline uses chained LLM passes, not a single prompt, to pull complete specs out of hardware datasheets.\n\nResearchers built an automated pipeline that turns hardware datasheets into hierarchical knowledge graphs. Instead of feeding a whole datasheet to an LLM in one shot, the system breaks documents into chunks, extracts properties from each piece, routes them into a hierarchy, and refines the graph iteratively while keeping a record of where each property came from. The team tested the approach on 20 datasheets covering different component types and document lengths. Against a one-shot baseline, the pipeline hit 98.8% property faithfulness versus 76.1%, and it captured 83.6% of everything the baseline got right, while the baseline captured only 10.5% of what the pipeline found.\n\nThe real finding here isn't that LLMs can read a datasheet. It's confirmation that dumping a whole document into a big context window and asking for all the specs quietly drops information, even when the document fits inside the window. That's a meaningful caveat for anyone building retrieval or search tools on top of long technical documents, not just hardware engineers.\n\nDatasheets are the kind of unglamorous, high-stakes text nobody wants to proofread by hand, which is exactly why automating it carefully, rather than quickly, is the right call.","[\"ai\",\"llm\",\"hardware\",\"knowledge-graphs\"]","2026-10-12T04:00:00.000Z","2026-10-12T04:31:50.279Z","2026-10-12T04:31:56.397Z","published",null,[],"ai",[24,26,27,28],"llm","hardware","knowledge-graphs",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12762",0,{"sections":35},[36,39,43,48,53,57,62,67,72,77,82,87,92,97],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",6896,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",946,{"name":44,"slug":45,"count":46,"latest_published_at":47},"Policy","policy",496,"2026-10-11T20:06:43.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",484,"2026-10-09T12:30:00.000Z",{"name":54,"slug":27,"count":55,"latest_published_at":56},"Hardware",234,"2026-10-11T17:20:00.000Z",{"name":58,"slug":59,"count":60,"latest_published_at":61},"Science","science",196,"2026-10-11T15:34:40.000Z",{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",190,"2026-10-11T21:14:40.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Startups","startups",120,"2026-10-09T17:02:31.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Software","software",115,"2026-10-11T07:34:45.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Dev Tools","dev-tools",110,"2026-10-09T13:03:48.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"General","general",70,"2026-10-11T14:15:00.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"Gaming","gaming",59,"2026-10-09T11:43:43.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]