# Do Latent Tokens Think? A Causal and Adversarial Analysis of Chain-of-Continuous-Thought

arXiv: https://arxiv.org/abs/2512.21711 · HTML: https://mlsanity.com/paper/2512.21711/

## Key facts

- Cited by 23 papers in the map (data through 14 Oct 2026).
- Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning.
- Builds on 12 of its 29 references that are in the map.
- First posted on arXiv 25 Dec 2025.

## Placed in

- LLM reasoning / [Prompting and chain-of-thought reasoning](https://mlsanity.com/lr/prompting-and-chain-of-thought-reasoning/)

## Builds on (most cited first, 12 of 12)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Training Large Language Models to Reason in a Continuous Latent Space](https://mlsanity.com/paper/2412.06769/) (2024)
- [Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thought](https://mlsanity.com/paper/2210.01240/) (2022)
- [Think before you speak: Training Language Models With Pause Tokens](https://mlsanity.com/paper/2310.02226/) (2023)
- [From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://mlsanity.com/paper/2405.14838/) (2024)
- [Implicit Chain of Thought Reasoning via Knowledge Distillation](https://mlsanity.com/paper/2311.01460/) (2023)
- [Implicit Reasoning in Large Language Models: A Comprehensive Survey](https://mlsanity.com/paper/2509.02350/) (2025)
- [Do Large Language Models Perform Latent Multi-Hop Reasoning without Exploiting Shortcuts?](https://mlsanity.com/paper/2411.16679/) (2024)
- [Editing Personality for Large Language Models](https://mlsanity.com/paper/2310.02168/) (2023)
- [Probing then Editing Response Personality of Large Language Models](https://mlsanity.com/paper/2504.10227/) (2025)
- [A Classification of Critical Configurations for any Number of Projective Views](https://mlsanity.com/paper/2401.03450/) (2024)

## Built on by (most cited first, 23 of 23)

- [The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook](https://mlsanity.com/paper/2604.02029/) (2026)
- [Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers](https://mlsanity.com/paper/2606.31779/) (2026)
- [Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Tasks](https://mlsanity.com/paper/2602.00449/) (2026)
- [Imagination Helps Visual Reasoning, But Not Yet in Latent Space](https://mlsanity.com/paper/2602.22766/) (2026)
- [Are Your Reasoning Models Reasoning or Guessing? A Mechanistic Analysis of Hierarchical Reasoning Models](https://mlsanity.com/paper/2601.10679/) (2026)
- [How Do Latent Reasoning Methods Perform Under Weak and Strong Supervision?](https://mlsanity.com/paper/2602.22441/) (2026)
- [What's Holding Back Latent Visual Reasoning?](https://mlsanity.com/paper/2605.18445/) (2026)
- [Chain Of Thought Compression: A Theoretical Analysis](https://mlsanity.com/paper/2601.21576/) (2026)
- [Beyond Visual Memory: Mechanistic Diagnostics of Latent Visual Reasoning](https://mlsanity.com/paper/2606.01287/) (2026)
- [Internalizing LLM Reasoning via Discovery and Replay of Latent Actions](https://mlsanity.com/paper/2602.04925/) (2026)
- [Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models](https://mlsanity.com/paper/2606.12689/) (2026)
- [What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis](https://mlsanity.com/paper/2606.20075/) (2026)
- [MUX: Continuous Reasoning via Multiplexed Tokens](https://mlsanity.com/paper/2607.18264/) (2026)
- [Opening the Black Box: A Survey on the Mechanisms of Multi-Step Reasoning in Large Language Models](https://mlsanity.com/paper/2601.14270/) (2026)
- [MemoSight: Unifying Context Compression and Multi Token Prediction for Reasoning Acceleration](https://mlsanity.com/paper/2604.14889/) (2026)
- [Robust and Efficient Guardrails with Latent Reasoning](https://mlsanity.com/paper/2605.29068/) (2026)
- [Final Checkpoints Are Not Enough: Analyzing Latent Reasoning Faithfulness Along Training Trajectories](https://mlsanity.com/paper/2607.06648/) (2026)
- [Assessing Adversarial Robustness of Latent Reasoning Models](https://mlsanity.com/paper/2609.22228/) (2026)
- [TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale](https://mlsanity.com/paper/2605.15053/) (2026)
- [HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps](https://mlsanity.com/paper/2606.03768/) (2026)
- [The Weight of Silence: A Causal Case for Weights Over the Scratchpad in Latent Chess Reasoning](https://mlsanity.com/paper/2607.20952/) (2026)
- [Dissecting Hierarchical Reasoning Models: A Mechanistic Study](https://mlsanity.com/paper/2609.22197/) (2026)
- [What Matters for Latent Reasoning with Flow Matching](https://mlsanity.com/paper/2610.06666/) (2026)

## Abstract (arXiv)

Latent tokens are gaining attention for enhancing reasoning in large language models (LLMs), yet their internal mechanisms remain unclear. This paper examines the problem from a reliability perspective, uncovering fundamental weaknesses: latent tokens function as uninterpretable placeholders rather than encoding faithful reasoning. While resistant to perturbation, they promote shortcut usage over genuine reasoning. We focus on Chain-of-Continuous-Thought (COCONUT), which claims better efficiency and stability than explicit Chain-of-Thought (CoT) while maintaining performance. We investigate this through two complementary approaches. First, steering experiments perturb specific token subsets, namely COCONUT and explicit CoT. Unlike CoT tokens, COCONUT tokens show minimal sensitivity to steering and lack reasoning-critical information. Second, shortcut experiments evaluate models under biased and out-of-distribution settings. Results on MMLU and HotpotQA demonstrate that COCONUT consistently exploits dataset artifacts, inflating benchmark performance without true reasoning. These findings reposition COCONUT as a pseudo-reasoning mechanism: it generates plausible traces that conceal shortcut dependence rather than faithfully representing reasoning processes.