SocialMaze: A Benchmark for Evaluating and Enhancing Social Reasoning in Large Language Models in Complex Social Environments
Paper • 2505.23713 • Published
Model Evaluation & Post-Training
Papers and datasets on social reasoning, cross-lingual weakness discovery, evaluator bias, and feedback for improving models.
-
-
Cross-Lingual Pitfalls: Automatic Probing Cross-Lingual Weakness of Multilingual Large Language Models
Paper • 2505.18673 • Published -
Inside the Unfair Judge: A Mechanistic Interpretability Account of LLM-as-Judge Bias
Paper • 2607.11871 • Published -
xzx34/SocialMaze
Viewer • Updated • 200k • 140 • 20 -
xzx34/cross-lingual-pitfalls
Viewer • Updated • 5.19k • 119