Eliciting Chain-of-Thought in Base LLMs via Gradient-Based Representation Optimization.

Zijian Wang, Yanxiang Ma, Chang Xu

Browse the full AAAI paper archive.