Coupled Variational Reinforcement Learning for Language Model General Reasoning 事件

Coupled Variational Reinforcement Learning for Language Model General Reasoning arXiv:2512.12576v3 Announce Type: replace Abstract: While reinforcement learning has achieved impressive progress in language model reasoning, it is constrained by the requirement for verifiable rewards. Recent verifier-free RL methods address this limitation by utilizing the probabilities that LLMs generate reference answers as reward signals. However, these approaches typically sample reasoning traces conditioned