Mar, 2024
私密基准测试以防止污染并提高对 LLM 的比较评估
Private Benchmarking to Prevent Contamination and Improve Comparative Evaluation of LLMs
Nishanth Chandran, Sunayana Sitaram, Divya Gupta, Rahul Sharma, Kashish Mittal...
TL;DR私密基准测试是解决基准测试数据被污染或泄露的问题的解决方案,并且可以保持模型的权重私密,以确保私密基准测试的高质量。