Steering LLMs? Actually, Sparse Autoencoders can outperform simple baselines 文章

ArXiv CS.CL2026-06-01NEWSen作者: Mikkel Godsk J{\o}rgensen, Lars Kai Hansen
查看原文 →
关系图谱

Steering LLMs? Actually, Sparse Autoencoders can outperform simple baselines · 相关技术

相关技术

Large language modelSparse Autoencoder (SAE)LoRA