PowerFlow: Unlocking the Dual Nature of LLMs via Principled Distribution Matching

ArXi:2603.18363v1 Announce Type: cross Unsupervised Reinforcement Learning from Internal Feedback (RLIF) has emerged as a promising paradigm for eliciting the latent capabilities of Large Language Models (LLMs) without external supervision. However, current methods rely on heuristic intrinsic rewards, which often lack a well-defined theoretical optimization target and are prone to degenerative biases. In this work, we