# World Model Robustness via Surprise Recognition

arXiv: https://arxiv.org/abs/2512.01119 · HTML: https://mlsanity.com/paper/2512.01119/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Latent dynamics models and model-based RL, one of 11 directions in World models.
- Builds on 20 of its 37 references that are in the map.
- First posted on arXiv 30 Nov 2025.

## Placed in

- World models / [Latent dynamics models and model-based RL](https://mlsanity.com/wm/latent-dynamics-models-and-model-based-rl/)

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

- [CARLA: An Open Urban Driving Simulator](https://mlsanity.com/paper/1711.03938/) (2017)
- [Restormer: Efficient Transformer for High-Resolution Image Restoration](https://mlsanity.com/paper/2111.09881/) (2021)
- [Recurrent World Models Facilitate Policy Evolution](https://mlsanity.com/paper/1809.01999/) (2018)
- [Leveraging Procedural Generation to Benchmark Reinforcement Learning](https://mlsanity.com/paper/1912.01588/) (2019)
- [Diffusion Models for Adversarial Purification](https://mlsanity.com/paper/2205.07460/) (2022)
- [Learning Invariant Representations for Reinforcement Learning without Reconstruction](https://mlsanity.com/paper/2006.10742/) (2020)
- [Benchmarking the Spectrum of Agent Capabilities](https://mlsanity.com/paper/2109.06780/) (2021)
- [Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark](https://mlsanity.com/paper/2310.12567/) (2023)
- [Robust Evaluation of Diffusion-Based Adversarial Purification](https://mlsanity.com/paper/2303.09051/) (2023)
- [SafeDreamer: Safe Reinforcement Learning with World Models](https://mlsanity.com/paper/2307.07176/) (2023)
- [Mask-based Latent Reconstruction for Reinforcement Learning](https://mlsanity.com/paper/2201.12096/) (2022)
- [Learning Latent Dynamic Robust Representations for World Models](https://mlsanity.com/paper/2405.06263/) (2024)
- [Learning End-to-end Multimodal Sensor Policies for Autonomous Navigation](https://mlsanity.com/paper/1705.10422/) (2017)
- [CarDreamer: Open-Source Learning Platform for World Model based Autonomous Driving](https://mlsanity.com/paper/2405.09111/) (2024)
- [The Role of Pretrained Representations for the OOD Generalization of Reinforcement Learning Agents](https://mlsanity.com/paper/2107.05686/) (2021)
- [MaDi: Learning to Mask Distractions for Generalization in Visual Deep Reinforcement Learning](https://mlsanity.com/paper/2312.15339/) (2023)
- [Novelty Detection in Reinforcement Learning with World Models](https://mlsanity.com/paper/2310.08731/) (2023)
- [LoRID: Low-Rank Iterative Diffusion for Adversarial Purification](https://mlsanity.com/paper/2409.08255/) (2024)
- [Policy-shaped prediction: avoiding distractions in model-based reinforcement learning](https://mlsanity.com/paper/2412.05766/) (2024)
- [Continual Reinforcement Learning via Autoencoder-Driven Task and New Environment Recognition](https://mlsanity.com/paper/2505.09003/) (2025)

## Built on by (most cited first, 4 of 4)

- [Safety in Embodied AI: A Survey of Risks, Attacks, and Defenses](https://mlsanity.com/paper/2605.02900/) (2026)
- [WMAttack: Automated Attack Search for Adversarial Evaluation of World-Model Agents](https://mlsanity.com/paper/2605.23220/) (2026)
- [Learning from Contrasts: Synthesizing Reasoning Paths from Diverse Search Trajectories](https://mlsanity.com/paper/2604.11365/) (2026)
- [ARB4WM: An Adversarial Robustness Benchmark for World Models in Continuous Control](https://mlsanity.com/paper/2606.16605/) (2026)

## Abstract (arXiv)

AI systems deployed in the real world must contend with distractions and out-of-distribution (OOD) noise that can destabilize their policies and lead to unsafe behavior. While robust training can reduce sensitivity to some forms of noise, it is infeasible to anticipate all possible OOD conditions. To mitigate this issue, we develop an algorithm that leverages a world model's inherent measure of surprise to reduce the impact of noise in world model--based reinforcement learning agents. We introduce both multi-representation and single-representation rejection sampling, enabling robustness to settings with multiple faulty sensors or a single faulty sensor. While the introduction of noise typically degrades agent performance, we show that our techniques preserve performance relative to baselines under varying types and levels of noise across multiple environments within self-driving simulation domains (CARLA and Safety Gymnasium). Furthermore, we demonstrate that our methods enhance the stability of two state-of-the-art world models with markedly different underlying architectures: Cosmos and DreamerV3. Together, these results highlight the robustness of our approach across world modeling domains. We release our code at https://github.com/Bluefin-Tuna/WISER .