Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning

• URL: http://arxiv.org/abs/2606.13106v1
• Date: Thu, 11 Jun 2026 09:33:40 GMT
• Title: Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning
• Abstract summary: We propose SWITCH, a switchable latent reasoning framework.<n>The model emits swi> to enter latent mode and /swi> to exit.<n>We show that SWITCH consistently outperforms prior hidden-state-recurrence latent reasoning approaches.
• Score: 38.056339518861506
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Latent chain-of-thought compresses reasoning by replacing visible reasoning traces with continuous hidden-state recurrence, but existing formulations are difficult to optimize with standard on-policy reinforcement learning (RL) and hard to interpret causally. Our key insight is that a single pair of explicit boundary tokens can address both issues at once: discrete entry and exit anchors make the latent block compatible with standard on-policy RL, and the same anchors offer a natural foothold for mechanistic analysis. Motivated by this, we propose SWITCH, a switchable latent reasoning framework. The model emits <swi> to enter latent mode and </swi> to exit. Because the boundaries are ordinary discrete tokens, the GRPO policy ratio is well-defined at every decision point. The same anchors also expose the latent steps to direct probing and causal intervention. We train the model with a visible-to-latent curriculum and a Switch-GRPO objective that propagates gradients through recurrent latent computation. SWITCH consistently outperforms prior hidden-state-recurrence latent reasoning approaches at similar scale. Mechanistic analysis through the boundary tokens further reveals three findings: (i) <swi> is a sharply localised, learned switching policy rather than a stylistic artefact; (ii) the latent step it opens performs problem-specific, causally important computation rather than acting as an inert placeholder; and (iii) that computation is concentrated at a single hidden-state transition on entry. Together, these results show that hidden-state-recurrence latent reasoning is both RL-trainable and open to direct mechanistic analysis, including of how on-policy RL itself improves the model from the inside.

Related papers

• SLPO: Scaling Latent Reasoning via a Surrogate Policy [25.465651511203337]
  Reinforcement learning with verifiable rewards has become the predominant recipe for eliciting test-time scaling in explicit Chain-of-Thought reasoners.<n>Latent reasoning instead carries intermediate computation as continuous vectors.<n>We introduce Surrogate Latent Policy Optimization to bring outcome-reward RL to autoregressive latent reasoners.
  arXiv  Detail & Related papers  (2026-07-22T02:45:08Z)
• Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models [47.195670444638715]
  We identify and formally define textbfEntropy-Gradient Inversion, a robust correlation between token entropy and logit gradients.<n>We propose textbfCorrelation-Regularized Group Policy Optimization (CorR-PO), which embeds this inversion signature into reinforcement learning.<n>Experiments on various reasoning benchmarks across multiple model scales show CorR-PO consistently outperforms state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-05-18T02:41:53Z)
• SAGE: Shaping Anchors for Guided Exploration in RLVR of LLMs [55.46289074417954]
  reinforcement learning with verifiable rewards (RLVR) reliably improves pass@1 on reasoning tasks, yet often fails to yield comparable gains in pass@k.<n>We argue that a central structural constraint arises from reverse-KL regularization, which stabilizes training but inherently anchors the policy to the reference distribution.<n>We propose SAGE, a principled framework that enables controllable empirical support expansion by reshaping the reverse-KL anchor distribution itself.
  arXiv  Detail & Related papers  (2026-05-15T07:42:21Z)
• Compliance versus Sensibility: On the Reasoning Controllability in Large Language Models [46.26628756478016]
  Large Language Models (LLMs) are known to acquire reasoning capabilities through shared inference patterns in pre-training data.<n>We show that reasoning conflicts are internally detectable, as confidence scores significantly drop during conflicting episodes.<n>Our findings establish that while LLM reasoning is anchored to concrete instances, active mechanistic interventions can effectively decouple logical schemata from data.
  arXiv  Detail & Related papers  (2026-04-29T22:55:40Z)
• Amortized Reasoning Tree Search: Decoupling Proposal and Decision in Large Language Models [2.5170433424424874]
  Reinforcement Learning with Verifiable Rewards has established itself as the dominant paradigm for instilling rigorous reasoning capabilities in Large Language Models.<n>We identify a critical pathology in this alignment process: the systematic suppression of valid but rare (low-likelihood under the base model distribution) reasoning paths.<n>We propose Amortized Reasoning Tree Search (ARTS) to counteract this collapse without discarding the base model's latent diversity.
  arXiv  Detail & Related papers  (2026-02-13T11:52:50Z)
• Towards Unsupervised Causal Representation Learning via Latent Additive Noise Model Causal Autoencoders [1.9732490977700972]
  Unsupervised representation learning seeks to recover latent generative factors.<n>Disentangling causal variables from observational data is impossible without supervision.<n>We propose the Latent Additive Noise Model Causal Autoencoder (LANCA) as a strong inductive bias for unsupervised discovery.
  arXiv  Detail & Related papers  (2025-12-15T10:52:30Z)
• Multi-Path Collaborative Reasoning via Reinforcement Learning [54.8518809800168]
  Chain-of-Thought (CoT) reasoning has significantly advanced the problem-solving capabilities of Large Language Models (LLMs)<n>Recent methods attempt to address this by generating soft abstract tokens to enable reasoning in a continuous semantic space.<n>We propose Multi-Path Perception Policy Optimization (M3PO), a novel reinforcement learning framework that explicitly injects collective insights into the reasoning process.
  arXiv  Detail & Related papers  (2025-12-01T10:05:46Z)
• CTRLS: Chain-of-Thought Reasoning via Latent State-Transition [57.51370433303236]
  Chain-of-thought (CoT) reasoning enables large language models to break down complex problems into interpretable intermediate steps.<n>We introduce groundingS, a framework that formulates CoT reasoning as a Markov decision process (MDP) with latent state transitions.<n>We show improvements in reasoning accuracy, diversity, and exploration efficiency across benchmark reasoning tasks.
  arXiv  Detail & Related papers  (2025-07-10T21:32:18Z)
• A Survey on Latent Reasoning [100.54120559169735]
  Large Language Models (LLMs) have demonstrated impressive reasoning capabilities.<n>CoT reasoning that verbalizes intermediate steps limits the model's expressive bandwidth.<n>Latent reasoning tackles this bottleneck by performing multi-step inference entirely in the model's continuous hidden state.
  arXiv  Detail & Related papers  (2025-07-08T17:29:07Z)
• Hybrid Latent Reasoning via Reinforcement Learning [50.6763762323985]
  We explore latent reasoning by leveraging the capabilities of large language models (LLMs) via reinforcement learning (RL)<n>We introduce hybrid reasoning policy optimization (HRPO), an RL-based hybrid latent reasoning approach that integrates prior hidden states into sampled tokens with a learnable gating mechanism.<n>HRPO-trained LLMs remain interpretable and exhibit intriguing behaviors like cross-lingual patterns and shorter completion lengths.
  arXiv  Detail & Related papers  (2025-05-24T01:26:16Z)

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.