[{"data":1,"prerenderedAt":36},["ShallowReactive",2],{"maker-remilouf":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":22,"domain":23,"tier":24,"attempts":25,"strip":26,"url":27,"mix":28,"praise":33,"objection":34,"sub":35},37125118,"Outlines","LLMs can generate valid JSON 100% of the time",false,"ai",2023,854,303,"curious",-0.11,1,[17,18,19,20,21],"constrained generation","llm json output","regex guided generation","pydantic llm","logit masking","remilouf","github.com","legend",0,"3300000314221111351335434220234443311011","https:\u002F\u002Fgithub.com\u002Fnormal-computing\u002Foutlines",[29,30,31,32],20,63,2,30,"Masking approach is efficient, with index built once and fast per-step lookups","Masking may distort the model's probability distribution and hurt output quality","constrained llm text generation",1794036794744]