# Thinking on the Fly: Test-Time Reasoning Enhancement via Latent Thought Policy Optimization

arXiv: https://arxiv.org/abs/2510.04182 · HTML: https://mlsanity.com/paper/2510.04182/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 16 of its 37 references that are in the map.
- First posted on arXiv 5 Oct 2025.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 16 of 16)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [Training Large Language Models to Reason in a Continuous Latent Space](https://mlsanity.com/paper/2412.06769/) (2024)
- [MathArena: Evaluating LLMs on Uncontaminated Math Competitions](https://mlsanity.com/paper/2505.23281/) (2025)
- [From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://mlsanity.com/paper/2405.14838/) (2024)
- [Deep Think with Confidence](https://mlsanity.com/paper/2508.15260/) (2025)
- [Scalable Best-of-N Selection for Large Language Models via Self-Certainty](https://mlsanity.com/paper/2502.18581/) (2025)
- [Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling](https://mlsanity.com/paper/2502.06703/) (2025)
- [Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space](https://mlsanity.com/paper/2505.13308/) (2025)

## Built on by (most cited first, 7 of 7)

- [The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook](https://mlsanity.com/paper/2604.02029/) (2026)
- [Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space](https://mlsanity.com/paper/2512.12623/) (2025)
- [How Do Latent Reasoning Methods Perform Under Weak and Strong Supervision?](https://mlsanity.com/paper/2602.22441/) (2026)
- [From Noise to Diversity: Random Embedding Injection in LLM Reasoning](https://mlsanity.com/paper/2605.11936/) (2026)
- [Right Direction, Wrong Step: Geometric Analysis of Finite-Step Failure in Looped Transformers](https://mlsanity.com/paper/2609.16665/) (2026)
- [GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning](https://mlsanity.com/paper/2608.02585/) (2026)
- [Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning](https://mlsanity.com/paper/2609.11393/) (2026)

## Abstract (arXiv)

Recent advancements in Large Language Models (LLMs) have shifted from explicit Chain-of-Thought (CoT) reasoning to more efficient latent reasoning, where intermediate thoughts are represented as vectors rather than text. However, latent reasoning can be brittle on challenging, out-of-distribution tasks where robust reasoning is most critical. To overcome these limitations, we introduce Latent Thought Policy Optimization (LTPO), a parameter-free framework that enhances LLM reasoning entirely at test time, without requiring model parameter updates. LTPO treats intermediate latent "thought" vectors as dynamic parameters that are actively optimized for each problem instance. It employs an online policy gradient method guided by an intrinsic, confidence-based reward signal computed directly from the frozen LLM's own output distributions, eliminating the need for external supervision or expensive text generation during optimization. Extensive experiments on five reasoning benchmarks show that LTPO not only matches or surpasses strong baselines on standard tasks but also demonstrates remarkable robustness where others fail. Most notably, on highly challenging AIME benchmarks where existing latent reasoning baselines collapse to near-zero accuracy, LTPO delivers substantial improvements, showcasing a unique capability for complex reasoning.