CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning

Researchers introduced CopT, a new LLM reasoning framework that drafts an answer first, then decides whether more thinking is needed.

CopT compares discrete and continuous-space estimates to gauge answer reliability and trigger selective reflection.

This training-free approach aims to reduce token use and latency while improving accuracy.

The method shows gains on math, coding, and agentic tasks, with efficiency benefits for large language models.