Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding
Source
arxiv.org
Author
Guang Yang, Homa Hosseinmardi, Fengchen Liu, Amir Ghasemian
Date
Why it matters
A single sycophancy score hides that task type drives yielding far more than model or pushback style. Designers can expect cheap-to-verify facts to hold and personal-choice or hard-to-verify claims to be endorsed.
articleLooking Again: Measuring Sycophancy in the Reasoning Chains of Multimodal Models Under PressureMahir Numayeer Islam, Gakuto Okuyama, Nikolaus Siauw, Shivank Garg, Madhur Panwar, Vasu Sharma
articleAgentic Scaffolding Amplifies Sycophantic Behavior in Large Language ModelsThantham Jittham
articleA Removal Based Approach to Improve LLM Faithfulness at Test-TimeQinglan Luo, S M A Nahian, John Guttag, S. Mazdak Abulnaga, Katie Matton