AI RESEARCH
Are Sparse Autoencoder Benchmarks Reliable?
arXiv CS.LG
•
ArXi:2605.18229v1 Announce Type: new Sparse autoencoders (SAEs) are a core interpretability tool for large language models, and progress on SAE architectures depends on benchmarks that reliably distinguish better SAEs from worse ones. We audit the SAE quality metrics in SAEBench, the de-facto standard SAE evaluation suite, through three complementary lenses: reseed noise on a fixed SAE, ground-truth correlation on synthetic SAEs, and discriminability across