Original post · AI translation
标题:“我们能让大语言模型变笨,还是它只是在装傻?一项关于可信的性能限制的实证研究”
这篇论文发现,基准测试的低分可以被精确制造出来,因此单凭低分几乎无法判断模型能做什么。
安全测试常将模型在危险能力测试中得低分视为令人放心的信号。但低分也可能是模型的表现受到了限制,因此单凭这个数字说明不了多少问题。
他们训练了一个 LoRA“旋钮”,使一个 Qwen 模型的表现趋近于较小的 Qwen 模型,然后将它的错误与真正的小模型的错误进行比较。他们还尝试通过提示让模型表现变差、提高温度参数,以及给权重加入噪声。
这个旋钮在未见过的测试中达到目标分数,误差不超过 1.9 分,而提示方法的误差为 42 至 47 分。但在多项选择题中,它新增的错误答案有多达 97% 集中在 1 个字母选项上;他们所考察的自然模型中,没有一个接近这一比例。
在将低分解读为能力低之前,应检查模型在哪些题目上出错,以及错误答案如何分布。
Original
This paper finds that a low benchmark score can be manufactured precisely, so a low score alone says little about what a model can do. Safety tests often read a low score on a dangerous-capability probe as reassurance. But a low score can come from a model being held back, so the number alone proves little. They trained a LoRA "dial" that pulls a Qwen model toward a smaller Qwen model, then compared its errors with the real small model's. They also tried prompting models to underperform, raising temperature, and adding weight noise. The dial hit unseen target scores within 1.9 points, while prompting missed by 42 to 47 points. But on multiple choice, up to 97% of its new errors landed on 1 letter, a share no natural model in their pool approached. Before reading a low score as low capability, check which items the model fails and how its wrong answers are spread.