Stabilizing Recurrent Dynamics for Test-Time Scalable Latent Reasoning in Looped Language Models 事件

Stabilizing Recurrent Dynamics for Test-Time Scalable Latent Reasoning in Looped Language Models arXiv:2605.26733v1 Announce Type: cross Abstract: Looped Language Models (LoopLMs) enable efficient latent reasoning through depth recurrence, yet exhibit unreliable test-time scaling behavior: performance often peaks at a certain iteration depth and then collapses with further recurrence. Through latent dynamics analysis, we find an inherent trade-off between stability and effectiveness in existing