# Performance Asymmetry in Model-Based Reinforcement Learning

arXiv: https://arxiv.org/abs/2505.19698 · HTML: https://mlsanity.com/paper/2505.19698/

## Key facts

- No citing papers in the map yet (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 58 references that are in the map.
- First posted on arXiv 26 May 2025.

## Placed in

- World models / [Latent dynamics models and model-based RL](https://mlsanity.com/wm/latent-dynamics-models-and-model-based-rl/)
- Deep reinforcement learning methods / [RL with diffusion and flow models](https://mlsanity.com/deeprl/rl-with-diffusion-and-flow-models/)

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

- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [Planning with Diffusion for Flexible Behavior Synthesis](https://mlsanity.com/paper/2205.09991/) (2022)
- [Mastering Atari with Discrete World Models](https://mlsanity.com/paper/2010.02193/) (2020)
- [Stochastic Interpolants: A Unifying Framework for Flows and Diffusions](https://mlsanity.com/paper/2303.08797/) (2023)
- [Revisiting Feature Prediction for Learning Visual Representations from Video](https://mlsanity.com/paper/2404.08471/) (2024)
- [TD-MPC2: Scalable, Robust World Models for Continuous Control](https://mlsanity.com/paper/2310.16828/) (2023)
- [Model-Based Reinforcement Learning for Atari](https://mlsanity.com/paper/1903.00374/) (2019)
- [Label-Efficient Semantic Segmentation with Diffusion Models](https://mlsanity.com/paper/2112.03126/) (2021)
- [Temporal Difference Learning for Model Predictive Control](https://mlsanity.com/paper/2203.04955/) (2022)
- [Transformers are Sample-Efficient World Models](https://mlsanity.com/paper/2209.00588/) (2022)
- [Transformer-based World Models Are Happy With 100k Interactions](https://mlsanity.com/paper/2303.07109/) (2023)
- [DiffuseVAE: Efficient, Controllable and High-Fidelity Generation from Low-Dimensional Latents](https://mlsanity.com/paper/2201.00308/) (2022)
- [Policy-Guided Diffusion](https://mlsanity.com/paper/2404.06356/) (2024)
- [Exploring Diffusion Time-steps for Unsupervised Representation Learning](https://mlsanity.com/paper/2401.11430/) (2024)
- [Diffusion World Model: Future Modeling Beyond Step-by-Step Rollout for Offline Reinforcement Learning](https://mlsanity.com/paper/2402.03570/) (2024)
- [DiffEnc: Variational Diffusion with a Learned Encoder](https://mlsanity.com/paper/2310.19789/) (2023)
- [MIRROR: Differentiable Deep Social Projection for Assistive Human-Robot Communication](https://mlsanity.com/paper/2203.02877/) (2022)

## Abstract (arXiv)

Recently, Model-Based Reinforcement Learning (MBRL) have achieved super-human level performance on the Atari100k benchmark on average. However, we discover that conventional aggregates mask a major problem, Performance Asymmetry: MBRL agents dramatically outperform humans in certain tasks (Agent-Optimal tasks) while drastically underperform humans in other tasks (Human-Optimal tasks). Indeed, despite achieving SOTA in the overall mean Human-Normalized Scores (HNS), the SOTA agent scored the worst among baselines on Human-Optimal tasks, with a striking 21X performance gap between the Human-Optimal and Agent-Optimal subsets. To address this, we partition Atari100k evenly into Human-Optimal and Agent-Optimal subsets, and introduce a more balanced aggregate, Sym-HNS. Furthermore, we trace the striking Performance Asymmetry in the SOTA pixel diffusion world model to the curse of dimensionality and its prowess on high visual detail tasks (e.g. Breakout). To this end, we propose a novel latent end-to-end Joint Embedding DIffusion (JEDI) world model that achieves SOTA results in Sym-HNS, Human-Optimal tasks, and Breakout -- thus reversing the worsening Performance Asymmetry trend while improving computational efficiency and remaining competitive on the full Atari100k.