Jul 2026
SLPO: Scaling Latent Reasoning via a Surrogate Policy
Surrogate Latent Policy Optimization (SLPO) is introduced to bring outcome-reward RL to autoregressive latent reasoners: an empirical surrogate policy density over latent transitions for trajectory-level credit assignment, and a correctness-supervised stopping head that outcome-reward optimization refines into a variable-horizon policy.
Runyang You, Zhiyuan Liu, Yongqi Li et al.
· arXiv.org · 1 citation