# DreamPlan: Efficient Reinforcement Fine-Tuning of Vision-Language Planners via Video World Models

arXiv: https://arxiv.org/abs/2603.16860 · HTML: https://mlsanity.com/paper/2603.16860/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and robot world models, one of 11 directions in World models.
- Builds on 18 of its 34 references that are in the map.
- First posted on arXiv 17 Mar 2026.

## Placed in

- World models / [Embodied and robot world models](https://mlsanity.com/wm/embodied-and-robot-world-models/)
- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- 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, 18 of 18)

- [Adding Conditional Control to Text-to-Image Diffusion Models](https://mlsanity.com/paper/2302.05543/) (2023)
- [CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer](https://mlsanity.com/paper/2408.06072/) (2024)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [Inner Monologue: Embodied Reasoning through Planning with Language Models](https://mlsanity.com/paper/2207.05608/) (2022)
- [Mastering Atari with Discrete World Models](https://mlsanity.com/paper/2010.02193/) (2020)
- [Genie: Generative Interactive Environments](https://mlsanity.com/paper/2402.15391/) (2024)
- [DayDreamer: World Models for Physical Robot Learning](https://mlsanity.com/paper/2206.14176/) (2022)
- [π*0.6: a VLA That Learns From Experience](https://mlsanity.com/paper/2511.14759/) (2025)
- [VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning](https://mlsanity.com/paper/2505.18719/) (2025)
- [Ctrl-World: A Controllable Generative World Model for Robot Manipulation](https://mlsanity.com/paper/2510.10125/) (2025)
- [WMPO: World Model-based Policy Optimization for Vision-Language-Action Models](https://mlsanity.com/paper/2511.09515/) (2025)
- [World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training](https://mlsanity.com/paper/2509.24948/) (2025)
- [VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World Model](https://mlsanity.com/paper/2602.12063/) (2026)
- [World-VLA-Loop: Closed-Loop Learning of Video World Model and VLA Policy](https://mlsanity.com/paper/2602.06508/) (2026)
- [Robot Learning from a Physical World Model](https://mlsanity.com/paper/2511.07416/) (2025)
- [AnchorDream: Repurposing Video Diffusion for Embodiment-Aware Robot Data Synthesis](https://mlsanity.com/paper/2512.11797/) (2025)

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

- [World Model for Robot Learning: A Comprehensive Survey](https://mlsanity.com/paper/2605.00080/) (2026)
- [World Action Planner: Generalizable Decision-Making with Action-Conditioned World Models](https://mlsanity.com/paper/2607.27599/) (2026)
- [World-Action Models for Robot Learning and Control: A Survey](https://mlsanity.com/paper/2609.16074/) (2026)
- [Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning](https://mlsanity.com/paper/2608.21204/) (2026)
- [WM-VLM: Probing Internal World Models for Interleaved Visual-Textual Reasoning](https://mlsanity.com/paper/2609.34826/) (2026)

## Abstract (arXiv)

Robotic manipulation requires sophisticated commonsense reasoning, a capability naturally possessed by large-scale Vision-Language Models (VLMs). While VLMs show promise as zero-shot planners, their lack of grounded physical understanding often leads to compounding errors and low success rates when deployed in complex real-world environments, particularly for challenging tasks like deformable object manipulation. Although Reinforcement Learning (RL) can adapt these planners to specific task dynamics, directly fine-tuning VLMs via real-world interaction is prohibitively expensive, unsafe, and sample-inefficient. To overcome this bottleneck, we introduce DreamPlan, a novel framework for the reinforcement fine-tuning of VLM planners via video world models. Instead of relying on costly physical rollouts, DreamPlan first leverages the zero-shot VLM to collect exploratory interaction data. We demonstrate that this sub-optimal data is sufficient to train an action-conditioned video generation model, which implicitly captures complex real-world physics. Subsequently, the VLM planner is fine-tuned entirely within the "imagination" of this video world model using Odds Ratio Policy Optimization (ORPO). By utilizing these virtual rollouts, physical and task-specific knowledge is efficiently injected into the VLM. Our results indicate that DreamPlan bridges the gap between semantic reasoning and physical grounding, significantly improving manipulation success rates without the need for large-scale real-world data collection. Our project page is https://psi-lab.ai/DreamPlan/.