NTLBench
收藏资源简介:
NTLBench是一个针对非迁移学习(NTL)的评估框架,由悉尼大学Sydney AI Centre的研究者创建。该框架统一了评估NTL性能和鲁棒性的标准,支持在9个数据集(超过116个域对)和5种网络架构家族上运行NTL和攻击方法,总计提供了至少40,000个实验配置。NTLBench的目的是为了推动鲁棒NTL方法的发展,并促进其在可信赖模型部署场景中的应用。
NTLBench is an evaluation framework for non-transfer learning (NTL), created by researchers from the Sydney AI Centre at The University of Sydney. This framework unifies the standard protocols for evaluating NTL performance and robustness. It supports executing NTL and attack methods across 9 datasets (over 116 domain pairs) and 5 families of network architectures, providing a total of at least 40,000 experimental configurations. The core objective of NTLBench is to advance the development of robust NTL methods and facilitate their application in trustworthy model deployment scenarios.

- 1Toward Robust Non-Transferable Learning: A Survey and Benchmark悉尼大学Sydney AI Centre · 2025年



