[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-model-as-judge":3},{"tag":4,"articles":9},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":8},"0da5d146-0add-4cbf-bed7-a4647e3cc502","model-as-judge",0,null,[10],{"id":11,"slug":12,"title":13,"summary":14,"category":15,"image_url":16,"cover_image":16,"language":17,"created_at":18},"d75b5708-d4ec-4c46-9592-fa0a68d4bc26","judge-reliability-harness-stress-tests-llm-judges-zh","LLM 評審也會不穩","這篇論文做了一個壓力測試工具，檢查 LLM 當評審時，會不會因為格式、改寫、篇幅或標籤翻轉而判斷不一致。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778740856189-g1zr.png","zh","2026-05-14T06:40:32.198872+00:00"]