# When Reasoning Helps Action: Monitoring and Steering Chain-of-Thought in Vision-Language-Action Policies

arXiv: https://arxiv.org/abs/2610.00601 · HTML: https://mlsanity.com/paper/2610.00601/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning.
- Builds on 22 of its 28 references that are in the map.
- First posted on arXiv 30 Sep 2026.

## Placed in

- LLM reasoning / [Prompting and chain-of-thought reasoning](https://mlsanity.com/lr/prompting-and-chain-of-thought-reasoning/)
- Vision-language-action models / [Reasoning and chain-of-thought in VLA](https://mlsanity.com/vla/reasoning-and-chain-of-thought-in-vla/)

## Builds on (most cited first, 22 of 22)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [A Survey on Multimodal Large Language Models](https://mlsanity.com/paper/2306.13549/) (2023)
- [CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models](https://mlsanity.com/paper/2503.22020/) (2025)
- [LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models](https://mlsanity.com/paper/2510.13626/) (2025)
- [PaliGemma 2: A Family of Versatile VLMs for Transfer](https://mlsanity.com/paper/2412.03555/) (2024)
- [Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail](https://mlsanity.com/paper/2511.00088/) (2025)
- [Controlled Decoding from Language Models](https://mlsanity.com/paper/2310.17022/) (2023)
- [Real-Time Anomaly Detection and Reactive Planning with Large Language Models](https://mlsanity.com/paper/2407.08735/) (2024)
- [Semantic Anomaly Detection with Large Language Models](https://mlsanity.com/paper/2305.11307/) (2023)
- [Value Augmented Sampling for Language Model Alignment and Personalization](https://mlsanity.com/paper/2405.06639/) (2024)
- [Do What You Say: Steering Vision-Language-Action Models via Runtime Reasoning-Action Alignment Verification](https://mlsanity.com/paper/2510.16281/) (2025)
- [DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Models](https://mlsanity.com/paper/2511.15669/) (2025)
- [PerceMon: Online Monitoring for Perception Systems](https://mlsanity.com/paper/2108.08289/) (2021)
- [Preemptive Detection and Steering of LLM Misalignment via Latent Reachability](https://mlsanity.com/paper/2509.21528/) (2025)
- [System-Level Safety Monitoring and Recovery for Perception Failures in Autonomous Vehicles](https://mlsanity.com/paper/2409.17630/) (2024)
- [Is VLA Reasoning Faithful? Probing Safety of Chain-of-Causation in Autonomous Driving Models](https://mlsanity.com/paper/2605.17268/) (2026)
- [LSRE: Latent Semantic Rule Encoding for Real-Time Semantic Risk Detection in Autonomous Driving](https://mlsanity.com/paper/2512.24712/) (2025)
- [VLADriveBench: Evaluating CoT-Action Relationship in VLA for Autonomous Driving](https://mlsanity.com/paper/2606.12706/) (2026)

## Abstract (arXiv)

Reasoning-enabled VLA policies expose chain-of-thought (CoT) traces that appear to explain and guide their actions, creating a potential interface for runtime safety through reasoning monitoring and correction. In this work, we define and operationalize two evaluation axes for assessing when this interface can improve embodied behavior: correctability, which measures whether unreliable reasoning can be detected and improved during generation, and actionability, which measures whether reasoning corrections produce behaviorally meaningful changes in the intended direction. To enable correctability, we introduce Token-level Reward for Utility-Steered Chain-of-Thought (TRUST), an offline-trained value model that predicts eventual reasoning correctness from partial prefixes and uses these estimates to monitor and selectively steer reasoning generation in frozen VLA policies. On the Alpamayo 1.5 driving VLA, TRUST monitors correctness with 88.9% accuracy and improves reasoning correctness from 75.9% to 90.0%. On a baseline-defined challenging subset in AlpaSim, TRUST reduces collision rate by 30.4% and maximum trajectory error by 11.5% relative to the unsteered policy, outperforming a compute-matched Best-of-4 baseline. On the DeepThinkVLA manipulation VLA, TRUST improves the correctness of grasp-state claims from 69.3% to 90.2% and action-choice claims from 68.8% to 85.9%, yet closed-loop task performance on LIBERO-Plus remains largely unchanged. Empirical analysis reveals intent-consistent behavioral effects in Alpamayo 1.5 but limited effects in DeepThinkVLA, helping interpret these different task-level outcomes. Together, our results show that gains in reasoning correctness do not automatically imply gains in embodied performance, motivating evaluation of correctability and actionability when using CoT as a runtime safety interface.