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

• URL: http://arxiv.org/abs/2605.26733v1
• Date: Tue, 26 May 2026 09:11:12 GMT
• Title: Stabilizing Recurrent Dynamics for Test-Time Scalable Latent Reasoning in Looped Language Models
• Abstract summary: Looped Language Models (LoopLMs) enable efficient latent reasoning through depth recurrence, yet exhibit unreliable test-time scaling behavior.<n>We propose STARS (ST-driven Recurrent Scaling), a training framework that constrains latent states to approach reductionally stable fixed points.
• Score: 24.301295080069476
• License: http://creativecommons.org/licenses/by/4.0/
• 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 architectures and strategies. By conceptualizing reasoning as uncertainty reduction, we propose that convergence toward stable fixed points while preserving effectiveness represents a promising way. To this end, we propose STARS (STAbility-driven Recurrent Scaling), a training framework that constrains latent states to approach asymptotically stable fixed points. This is realized via efficient Jacobian Spectral Radius Regularization with random loop sampling, enabling STARS to maximize effectiveness while ensuring rigorous stability. Experiments on arithmetic tasks show that STARS achieves reliable test-time scaling, and on complex mathematical reasoning it substantially mitigates performance degradation as recurrence depth increases while also improving peak performance.

Related papers

• SGD at the Edge of Stability: Stochastic Stabilization with Large Learning Rates [2.104588729923885]
  Modern deep learning has been shown to operate at the edge of stability.<n>We provide convergence guarantees for Gradient Descent (SGD) applied to the multiclass cross-entropy loss.<n> Experiments validate our theoretical findings and illustrate the benefits of SGD in the large-stepsize regime.
  arXiv  Detail & Related papers  (2026-06-29T21:32:58Z)
• Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning [81.57028614960576]
  We show that the Confidence Region possesses two critical properties: 1) High Reliability -- answers in the confidence region become highly accurate and stable, and 2) High Redundancy -- models generate unnecessary tokens long after reaching the correct answer.<n>These properties unlock more efficient and reliable inference strategies.
  arXiv  Detail & Related papers  (2026-06-01T10:11:14Z)
• Think Twice Before You Write -- an Entropy-based Decoding Strategy to Enhance LLM Reasoning [32.332197731504046]
  Decoding strategies play a central role in shaping the reasoning ability of large language models.<n>Traditional methods such as greedy decoding and beam search often suffer from error propagation.<n>We propose an entropy-guided decoding framework that introduces token-level adaptivity into generation.
  arXiv  Detail & Related papers  (2026-03-10T23:08:26Z)
• Fast Rerandomization for Balancing Covariates in Randomized Experiments: A Metropolis-Hastings Framework [12.297315057225532]
  We propose an algorithm that achieves speedups ranging from 10 to 10,000 times while maintaining exact and validity.<n>Our proposed algorithm, PSRSRR, achieves speedups ranging from 10 to 10,000 times while maintaining exact and validity.
  arXiv  Detail & Related papers  (2026-02-07T16:40:27Z)
• Not All Preferences Are Created Equal: Stability-Aware and Gradient-Efficient Alignment for Reasoning Models [52.48582333951919]
  We propose a dynamic framework designed to enhance alignment reliability by maximizing the Signal-to-Noise Ratio of policy updates.<n>SAGE (Stability-Aware Gradient Efficiency) integrates a coarse-grained curriculum mechanism that refreshes candidate pools based on model competence.<n> Experiments on multiple mathematical reasoning benchmarks demonstrate that SAGE significantly accelerates convergence and outperforms static baselines.
  arXiv  Detail & Related papers  (2026-02-01T12:56:10Z)
• Adaptive Variance-Penalized Continual Learning with Fisher Regularization [0.0]
  This work presents a novel continual learning framework that integrates Fisher-weighted asymmetric regularization of parameter variances.<n>Our method dynamically modulates regularization intensity according to parameter uncertainty, achieving enhanced stability and performance.
  arXiv  Detail & Related papers  (2025-08-15T21:49:28Z)
• Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space [92.6187727249868]
  We introduce LatentSeek, a framework that enhances reasoning through Test-Time Instance-level Adaptation (TTIA) within the model's latent space.<n>LatentSeek is evaluated on a range of reasoning benchmarks, including GSM8K, MATH-500, and AIME2024.<n>Results show that LatentSeek consistently outperforms strong baselines.
  arXiv  Detail & Related papers  (2025-05-19T16:26:02Z)
• Robust Reinforcement Learning under Diffusion Models for Data with Jumps [40.2559197706778]
  We introduce the Mean-Square Bipower Variation Error (MSBVE) algorithm, which enhances robustness and convergence in scenarios involving significant noise and jumps.<n>We first revisit the Mean-Square TD Error (MSTDE) algorithm, commonly used in continuous-time RL, and highlight its limitations in handling jumps in state dynamics.<n>The proposed MSBVE algorithm minimizes the mean-square quadratic variation error, offering improved performance over MSTDE in environments characterized by SDEs with jumps.
  arXiv  Detail & Related papers  (2024-11-18T16:17:34Z)
• Training Generative Adversarial Networks by Solving Ordinary Differential Equations [54.23691425062034]
  We study the continuous-time dynamics induced by GAN training. From this perspective, we hypothesise that instabilities in training GANs arise from the integration error. We experimentally verify that well-known ODE solvers (such as Runge-Kutta) can stabilise training.
  arXiv  Detail & Related papers  (2020-10-28T15:23:49Z)
• Fine-Grained Analysis of Stability and Generalization for Stochastic Gradient Descent [55.85456985750134]
  We introduce a new stability measure called on-average model stability, for which we develop novel bounds controlled by the risks of SGD iterates. This yields generalization bounds depending on the behavior of the best model, and leads to the first-ever-known fast bounds in the low-noise setting. To our best knowledge, this gives the firstever-known stability and generalization for SGD with even non-differentiable loss functions.
  arXiv  Detail & Related papers  (2020-06-15T06:30:19Z)
• Distributional Robustness and Regularization in Reinforcement Learning [62.23012916708608]
  We introduce a new regularizer for empirical value functions and show that it lower bounds the Wasserstein distributionally robust value function. It suggests using regularization as a practical tool for dealing with $textitexternal uncertainty$ in reinforcement learning.
  arXiv  Detail & Related papers  (2020-03-05T19:56:23Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.