LL

LLM 当裁判 (LLM as a judge)

概念
本站收录 11 集 · 3 条金句 · 关联 10 个

集里怎么说它

① 提到它的金句

3 条

我认为有很多它表现不太好的情况,但我觉得如果你把 LLM 当裁判的用途从一个基准重新定义为一个异常检测器,那它其实是可以的。
Well, I think there’s a lot of cases where it doesn’t work very well, but I think if you reframe the utility of an LLM as a judge from being a benchmark to being an anomaly detector, then I think it’s actually okay.
—— Ankur Goyal · [25:10]

指向原始笔记的链接

顺便说一句,这比大多数 LLM-as-a-judge 方法——也就是让一个 LLM 来判断那是优质的人类作品还是 AI 生成的 slop——表现都要好。
This performed better, by the way, than most LLM-as-a-judge methods of asking an LLM to judge if that is great human quality versus AI-generated slop.
—— Thais Castello Branco · [08:29]

指向原始笔记的链接

他们开始以某种方式利用测试数据,用 LLM 作为裁判来训练模型,用各种花招来刷基准,拿季度奖金之类的。
They start to kind of use leverage test data to use LLM as a judge for training of the models, use all the various tricks to basically gain benchmarks, get like quarterly bonuses and so on.
—— Alex Mashrabov · [23:52]

指向原始笔记的链接

② 出现在这些集

11 集

③ 关联

点进去有真内容 —— 本页主要出口

智能体 · 评估 · Claude · Codex · Anthropic · OpenAI · 护栏 · 错误分析 · harness · GitHub