CorrSteer: Steering Improves Task Performance and Safety in LLMs through Correlation-based Sparse Autoencoder Feature Selection Paper โข 2508.12535 โข Published Aug 18, 2025 โข 2
Running 52 Bringing paper to life: A modern template for scientific writing ๐ 52 Download a ready-to-use scientific paper template
FaithfulSAE: Towards Capturing Faithful Features with Sparse Autoencoders without External Dataset Dependencies Paper โข 2506.17673 โข Published Jun 21, 2025 โข 7