On the Intrinsic Self-Correction Capability of LLMs: Uncertainty and Latent Concept

Open in new window