How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs? 文章

ArXiv CS.CL2026-07-21PAPERen作者: Prakhar Gupta, Terry Jingchen Zhang, Florent Draye, Bernhard Sch\"olkopf, Zhijing Jin

查看原文 →

How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs? · 相关技术

相关技术

pretraining sycophancy hidden states causal intervention leave-one-dataset-out transfer Language model probing alignment tuning CBCT cue-induced biases LLM