tracked author

Sergey Levine

VIMPO coauthor. UC Berkeley EECS homepage identifies him as an associate professor, RAIL Lab researcher, and deep reinforcement learning / robotics / decision-making scholar; homepage directly links the @svlevine Twitter/X account.

1 archived notes X: high Homepage X

Related Notes

按论文归档时间排序，展示该作者在本站已经出现的材料。

Jun 21, 2026

VIMPO: Value Implicit Policy Optimization for LLMs

VIMPO 试图填补 GRPO 和 PPO actor critic 之间的空位：它不训练独立 critic，却从 KL regularized RL 的最优性条件推出一个由 policy reference log ratio 表达的隐式 value recurrence，用终止状态 $V(s T)=0$ 把 outcome reward 变成 value loss，再用同一个 log ratio identity 构造 token...

2606.20008-vimpo-value-implicit-policy-optimization-llms RL Algorithm Credit Assignment Reasoning RL