[{"data":1,"prerenderedAt":32},["ShallowReactive",2],{"maker-xlayn":3},[4],{"id":5,"name":6,"line":7,"ai":8,"cat":9,"year":10,"points":11,"comments":12,"rec":13,"s":14,"top":15,"kw":16,"maker":21,"domain":22,"tier":23,"attempts":24,"strip":25,"url":26,"mix":27,"praise":30,"objection":31,"sub":17},47431671,"llm-circuit-finder","Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training",false,"ai",2026,265,80,"curious",-0.12,5,[17,18,19,20],"llm layer duplication","reasoning circuits","transformer layers","lm-eval benchmarks","xlayn","github.com","legend",0,"1000100105310333353335033311300110533513","https:\u002F\u002Fgithub.com\u002Falainnothere\u002Fllm-circuit-finder",[15,28,24,29],27,9,"Interesting, cool result on the logical deduction benchmark","Prior art: David Ng's RYS method and earlier layer duplication work",1793991318129]