[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-instruction-following":3},{"tag":4,"articles":10},{"id":5,"name":6,"slug":7,"article_count":8,"description_zh":9,"description_en":9},"dda0af4d-0991-458d-920a-dfb1c40acb7b","instruction following","instruction-following",0,null,[11],{"id":12,"slug":13,"title":14,"summary":15,"category":16,"image_url":17,"cover_image":17,"language":18,"created_at":19},"140a1bc8-8432-4950-9ed7-f28ea3060068","llms-procedural-execution-diagnostic-study-zh","LLM 會算，但不一定照步驟做","這篇診斷研究直接測 LLM 能不能照程序一步一步執行。結果顯示，步驟一拉長，模型的程序忠實度就明顯下滑，算術本身卻不難。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777875651857-35bu.png","zh","2026-05-04T06:20:26.283075+00:00"]