# Emergent World Representations in OpenVLA

arXiv: https://arxiv.org/abs/2509.24559 · HTML: https://mlsanity.com/paper/2509.24559/

## Key facts

- Cited by 6 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 20 of its 38 references that are in the map.
- First posted on arXiv 29 Sep 2025.

## Placed in

- World models / [Embodied and robot world models](https://mlsanity.com/wm/embodied-and-robot-world-models/)
- Vision-language-action models / [VLA architectures and generalization](https://mlsanity.com/vla/vla-architectures-and-generalization/)
- Mechanistic interpretability / [Steering and probing for LLM safety](https://mlsanity.com/mechinterp/steering-and-probing-for-llm-safety/)
- Deep reinforcement learning methods / [RL for robot locomotion and manipulation](https://mlsanity.com/deeprl/rl-for-robot-locomotion-and-manipulation/)

## Builds on (most cited first, 20 of 20)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Efficient Estimation of Word Representations in Vector Space](https://mlsanity.com/paper/1301.3781/) (2013)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems](https://mlsanity.com/paper/2005.01643/) (2020)
- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model](https://mlsanity.com/paper/1911.08265/) (2019)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](https://mlsanity.com/paper/2403.19647/) (2024)
- [Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task](https://mlsanity.com/paper/2210.13382/) (2022)
- [Emergent Linear Representations in World Models of Self-Supervised Sequence Models](https://mlsanity.com/paper/2309.00941/) (2023)
- [Scaling Laws for Transfer](https://mlsanity.com/paper/2102.01293/) (2021)
- [Benchmarking Model-Based Reinforcement Learning](https://mlsanity.com/paper/1907.02057/) (2019)
- [On Convergence of Extended Dynamic Mode Decomposition to the Koopman Operator](https://mlsanity.com/paper/1703.04680/) (2017)
- [Interpreting CLIP with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2502.20578/) (2025)
- [Emergence of Maps in the Memories of Blind Navigation Agents](https://mlsanity.com/paper/2301.13261/) (2023)
- [Embedding Arithmetic of Multimodal Queries for Image Retrieval](https://mlsanity.com/paper/2112.03162/) (2021)
- [Group Generalized Mean Pooling for Vision Transformer](https://mlsanity.com/paper/2212.04114/) (2022)

## Built on by (most cited first, 6 of 6)

- [Observing and Controlling Features in Vision-Language-Action Models](https://mlsanity.com/paper/2603.05487/) (2026)
- [Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models](https://mlsanity.com/paper/2603.19183/) (2026)
- [Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models](https://mlsanity.com/paper/2603.19233/) (2026)
- [Decoding Task Progress from VLA Representations](https://mlsanity.com/paper/2608.13474/) (2026)
- [Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models](https://mlsanity.com/paper/2605.00321/) (2026)
- [Present but Not Remembered: Auditing How Frozen VLAs Encode, Deploy, and Steer Visual History](https://mlsanity.com/paper/2607.03372/) (2026)

## Abstract (arXiv)

Vision Language Action models (VLAs) trained with policy-based reinforcement learning (RL) encode complex behaviors without explicitly modeling environmental dynamics. However, it remains unclear whether VLAs implicitly learn world models, a hallmark of model-based RL. We propose an experimental methodology using embedding arithmetic on state representations to probe whether OpenVLA, the current state of the art in VLAs, contains latent knowledge of state transitions. Specifically, we measure the difference between embeddings of sequential environment states and test whether this transition vector is recoverable from intermediate model activations. Using linear and non linear probes trained on activations across layers, we find statistically significant predictive ability on state transitions exceeding baselines (embeddings), indicating that OpenVLA encodes an internal world model (as opposed to the probes learning the state transitions). We investigate the predictive ability of an earlier checkpoint of OpenVLA, and uncover hints that the world model emerges as training progresses. Finally, we outline a pipeline leveraging Sparse Autoencoders (SAEs) to analyze OpenVLA's world model.