You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This repository contains the dataset and code used in our paper, “I Am Aligned, But With Whom? Diagnosing Structural Alignment Failures in Multilingual LLMs” It provides tools to evaluate how LLMs represent cultural values across 16 countries, multiple languages, and perspectives.
Cross-domain limits of hand-crafted CoT-surface features: AUROC 0.982 in math, 0.434 in coding. Five methods, one conclusion—code correctness is not in the text.