mlsanityData to 14 Oct 2026

RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step

arXiv 2605.09346 ↗First posted 10 May 2026cs.CL cs.AIcited by 0

On the map: LLM reasoning → Test-time compute and efficient thinking

also in 1 more
Abstract (arXiv)

The Chain-of-Thought (CoT) paradigm, while enhancing the interpretability of Large Language Models (LLMs), is constrained by the inefficiencies and expressive limits of natural language. Latent Chain-of-Thought (latent CoT) reasoning, which operates in a continuous latent space, offers a promising alternative but faces challenges from structural complexities in existing multi-step or multi-model paradigms, such as error propagation and coordination overhead. In this paper, we introduce One-Model One-Step, a novel compression framework for Latent Reasoning with Rule-Based Priors(RuPLaR) to address this challenge. Our method trains an LLM to autonomously generate latent reasoning tokens in a single training stage, guided by rule-based prior probability distributions, thereby eliminating cascaded processes and inter-model dependencies. To ensure reasoning quality, we design a joint training objective that enforces answer consistency via cross-entropy, aligns soft tokens with rule-based priors via KL divergence (the Soft Thinking constraint), and adds a problem-thought semantic alignment constraint in the representation space. Extensive experiments show that our compression framework not only improves accuracy by 11.1% over existing latent CoT methods but also achieves this with minimal token usage, underscoring its effectiveness and extensibility. Code: https://github.com/xiaocen-luo/RuPLaR.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 18 of its 41 references that are in the map. First posted on arXiv 10 May 2026.

Builds on

18 of 41 references.

Closest ancestors · 6

Same direction, newest first

  1. 2026ReGuLaR: Variational Latent Reasoning Guided by Rendered Chain-of-Thought30 Jan
  2. 2025LatentEvolve: Self-Evolving Test-Time Scaling in Latent Space29 Sep
  3. 2025Fast Thinking for Large Language Models28 Sep
  4. 2025LTA-thinker: Latent Thought-Augmented Training Framework for Large Language Models on Complex Reasoning16 Sep
  5. 2025SoftCoT++: Test-Time Scaling with Soft Chain-of-Thought Reasoning16 May
  6. 2025Reasoning to Learn from Latent Thoughts24 Mar

Other references · 12

  1. 2023GPT-4 Technical Reportcited by 18,396
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  5. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 384
  6. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 235
  7. 2023Implicit Chain of Thought Reasoning via Knowledge Distillationcited by 176
  8. 2025Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoningcited by 78
  9. 2025SofT-GRPO: Surpassing Discrete-Token LLM Reinforcement Learning via Gumbel-Reparameterized Soft-Thinking Policy Optimizationcited by 21
  10. 2025Enhancing Latent Computation in Transformers with Latent Tokenscited by 14
  11. 2025SynAdapt: Learning Adaptive Reasoning in Large Language Models via Synthetic Continuous Chain-of-Thoughtcited by 13
  12. 2025Deep Thinking by Markov Chain of Continuous Thoughtscited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains
  2. 2025Think Silently, Think Fast: Dynamic Latent Compression of LLM Reasoning Chains
  3. 2026SYNLAT: Syntax-Aligned Text-Latent Compression for Chain-of-Thought Reasoning
  4. 2026Chain Of Thought Compression: A Theoretical Analysis
  5. 2026Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression
  6. 2026Compress the Easy, Explore the Hard: Difficulty-Aware Entropy Regularization for Efficient LLM Reasoning
  7. 2026A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM
  8. 2026Latent Thought Flow: Efficient Latent Reasoning in Large Language Models