# VLCP: Vision Language Control Policy Closed-Loop Code Replanning for Robot Manipulation

arXiv: https://arxiv.org/abs/2608.16978 · HTML: https://mlsanity.com/paper/2608.16978/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 12 of its 13 references that are in the map.
- First posted on arXiv 17 Aug 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

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

- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [π0: A Vision-Language-Action Flow Model for General Robot Control](https://mlsanity.com/paper/2410.24164/) (2024)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [Code as Policies: Language Model Programs for Embodied Control](https://mlsanity.com/paper/2209.07753/) (2022)
- [Inner Monologue: Embodied Reasoning through Planning with Language Models](https://mlsanity.com/paper/2207.05608/) (2022)
- [ProgPrompt: Generating Situated Robot Task Plans using Large Language Models](https://mlsanity.com/paper/2209.11302/) (2022)
- [CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation](https://mlsanity.com/paper/2603.22435/) (2026)
- [Code-as-Monitor: Constraint-aware Visual Programming for Reactive and Proactive Robotic Failure Detection](https://mlsanity.com/paper/2412.04455/) (2024)
- [DoReMi: Grounding Language Model by Detecting and Recovering from Plan-Execution Misalignment](https://mlsanity.com/paper/2307.00329/) (2023)
- [VLA-0: Building State-of-the-Art VLAs with Zero Modification](https://mlsanity.com/paper/2510.13054/) (2025)
- [Robotic Programmer: Video Instructed Policy Code Generation for Robotic Manipulation](https://mlsanity.com/paper/2501.04268/) (2025)
- [Act-Observe-Rewrite: Multimodal Coding Agents as In-Context Policy Learners for Robot Manipulation](https://mlsanity.com/paper/2603.04466/) (2026)

## Built on by (most cited first, 5 of 5)

- [Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens](https://mlsanity.com/paper/2610.01939/) (2026)
- [Embodied Turing Machines: Stateful Code for Robot Recursive Self-Improvement](https://mlsanity.com/paper/2610.12369/) (2026)
- [CAPEX: Efficiently Distilling Foundation Model Behavior into Deployable Robot Policies through Experience-Adaptive Reasoning](https://mlsanity.com/paper/2609.33007/) (2026)
- [CodeActionBench: Evaluating Agentic Code-as-Policy for Embodied Manipulation](https://mlsanity.com/paper/2609.33807/) (2026)
- [RoboAware: Learning to Coordinate Embodied Skills from Counterfactual Outcomes](https://mlsanity.com/paper/2610.11480/) (2026)

## Abstract (arXiv)

Turning a frontier vision-language model into a robot policy usually means fine-tuning it to emit an action representation it never saw in pretraining, which throws away much of the reasoning that made the model worth reaching for. We go the other way and keep the VLM frozen. It writes the policy as a short Python control function, with no demonstrations and no fine-tuning. Writing that code once is open-loop, though. Existing closed-loop methods react at the wrong level: they retry a fixed policy or pick a different subtask, but never rewrite the code that failed. VLCP closes the loop where the failure actually lives, on the control code, within a single episode. Every K steps the VLM re-observes the scene from multi-view RGB, proprioceptive state, and a state delta, then rewrites the control function from what it just saw, so a failure is caught before it compounds. We evaluate on a 57-task MuJoCo/RoboVerse sweep. This training-free policy reaches 35.1% pooled success, against 3.5% for the identical system queried once per episode. That tenfold gap holds with non-overlapping confidence intervals in every scene family. The gain traces to a 27.3% within-episode recovery rate on failed grasps: a miss an open-loop controller would carry to the end of the episode gets re-observed and fixed at the next replan. And the loop stays cheap. A median 84% of input tokens hit cache, an episode needs only about 10 compact queries, and control blocks written during any replan persist to a cross-episode skill library reused in later prompts.