mlsanityData to 7 Oct 2026

A Formal Comparison Between Chain of Thought and Latent Thought

arXiv 2509.25239 ↗First posted 25 Sep 2025cs.AI cs.CL cs.LGcited by 14

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

Abstract (arXiv)

Chain of thought (CoT) elicits reasoning in large language models by explicitly generating intermediate tokens. In contrast, latent thought reasoning operates directly in the continuous latent space, enabling computation beyond discrete linguistic representations. While both approaches exploit iterative computation, their comparative capabilities remain underexplored. In this work, we present a formal analysis showing that latent thought admits more efficient parallel computation than inherently sequential CoT. In contrast, CoT enables approximate counting and sampling through stochastic decoding. These separations suggest the tasks for which depth-driven recursion is more suitable, thereby offering practical guidance for choosing between reasoning paradigms.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 32 of its 58 references that are in the map. First posted on arXiv 25 Sep 2025.

Builds on

32 of 58 references.

Closest ancestors · 11

Same direction, newest first

  1. 2025Continuous Chain of Thought Enables Parallel Exploration and Reasoning29 May
  2. 2025Exact Expressive Power of Transformers with Padding25 May
  3. 2025Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought18 May
  4. 2025Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers4 Feb
  5. 2024Training Large Language Models to Reason in a Continuous Latent Space9 Dec
  6. 2024On the Representational Capacity of Neural Language Models with Chain-of-Thought Reasoning20 Jun
  7. 2024Understanding Transformer Reasoning Capabilities via Graph Algorithms28 May
  8. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step23 May
  9. 2024Chain of Thought Empowers Transformers to Solve Inherently Serial Problems20 Feb
  10. 2023Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspective24 May
  11. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Models28 Jan

Other references · 21

  1. 2017Attention Is All You Needcited by 59,932
  2. 2019ALBERT: A Lite BERT for Self-supervised Learning of Language Representationscited by 2,772
  3. 2018Universal Transformerscited by 613
  4. 2023Looped Transformers as Programmable Computerscited by 224
  5. 2025Reasoning with Latent Thoughts: On the Power of Looped Transformerscited by 223
  6. 2022The Parallelism Tradeoff: Limitations of Log-Precision Transformerscited by 191
  7. 2025Hierarchical Reasoning Modelcited by 167
  8. 2021Saturated Transformers are Constant-Depth Threshold Circuitscited by 137
  9. 2023Looped Transformers are Better at Learning Learning Algorithmscited by 135
  10. 2025Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computationcited by 120
  11. 2024Looped Transformers for Length Generalizationcited by 91
  12. 2024Transformers, parallel computation, and logarithmic depthcited by 75
9 more
  1. 2025Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinkingcited by 56
  2. 2024MoEUT: Mixture-of-Experts Universal Transformerscited by 55
  3. 2025A Little Depth Goes a Long Way: The Expressive Power of Log-Depth Transformerscited by 49
  4. 2024Artificial Kuramoto Oscillatory Neuronscited by 37
  5. 2024Learning Iterative Reasoning through Energy Diffusioncited by 36
  6. 2024Simulation of Graph Algorithms with Looped Transformerscited by 29
  7. 2024On Expressive Power of Looped Transformers: Theoretical Analysis and Enhancement via Timestep Encodingcited by 28
  8. 2025Continuous Thought Machinescited by 21
  9. 2024Looped ReLU MLPs May Be All You Need as Practical Programmable Computerscited by 19

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 14.

14 papers.

Sort:

By area:

  1. 2026Solve the Loop: Attractor Models for Language and Reasoningcited by 11
  2. 2026Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformerscited by 11
  3. 2026Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Taskscited by 10
  4. 2026Are Your Reasoning Models Reasoning or Guessing? A Mechanistic Analysis of Hierarchical Reasoning Modelscited by 9
  5. 2026Residual Context Diffusion Language Modelscited by 9
  6. 2026How Do Latent Reasoning Methods Perform Under Weak and Strong Supervision?cited by 9
  7. 2026Chain-of-Thought and Compressed Looped Transformers: A Memory-Budget Separationcited by 2
  8. 2026What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysiscited by 2
  9. 2026Training Continuous Chain of Thought Models: A Tale of Two Regimescited by 2
  10. 2026Exponential Sample Complexity Separation between Flat and Hierarchical Agentic Theorem Proverscited by 0
  11. 2026Recursive Scaling in Masked Diffusion Modelscited by 0
  12. 2026Soft Token Alignment for Cross-Lingual Reasoningcited by 0
2 more
  1. 2026BDH-CQ: In-Context Learning with Recurrent Latent Reasoningcited by 0
  2. 2026Looped Diffusion Transformercited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning
  2. 2025Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer
  3. 2025Do Latent Tokens Think? A Causal and Adversarial Analysis of Chain-of-Continuous-Thought
  4. 2026Latent Chain-of-Thought as Planning: Decoupling Reasoning from Verbalization
  5. 2025Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones
  6. 2026Latent Reasoning with Supervised Thinking States
  7. 2026Latent Reasoning with Normalizing Flows
  8. 2026Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought