3 papers
cs.LG2026
Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing
Parsa Mirtaheri, Mikhail Belkin
Large language models (LLMs) can produce chains of thought (CoT) that do not accurately reflect the actual factors driving their answers. In multiple-choice settings with an inject…
cs.LG2026
Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones
Parsa Mirtaheri, Ezra Edelman, Samy Jelassi +2
Inference-time computation has emerged as a promising scaling axis for improving large language model reasoning. However, despite yielding impressive performance, the optimal alloc…
cs.AI2025
Direct Alignment with Heterogeneous Preferences
Ali Shirali, Arash Nasr-Esfahany, Abdullah Alomar +3
Alignment with human preferences is commonly framed using a universal reward function, even though human preferences are inherently heterogeneous. We formalize this heterogeneity b…