AIBearisharXiv – CS AI · 5h ago7/10
🧠
Fragile Preferences: A Deep Dive Into Order Effects in Large Language Models
Researchers conducted the first systematic study of order bias in Large Language Models used for high-stakes decision-making, finding that LLMs exhibit strong position effects and previously undocumented name biases that can lead to selection of strictly inferior options. The study reveals distinct failure modes in AI decision-support systems, with proposed mitigation strategies using temperature parameter adjustments to recover underlying preferences.