OneBench
How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs? | OneBench: AI Insights