# Stealing That Free Lunch: Exposing the Limits of Dyna-Style Reinforcement Learning

arXiv: https://arxiv.org/abs/2412.14312 · HTML: https://mlsanity.com/paper/2412.14312/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Deep RL algorithms and representations, one of 15 directions in Deep reinforcement learning methods.
- Builds on 35 of its 55 references that are in the map.
- First posted on arXiv 18 Dec 2024.

## Placed in

- Deep reinforcement learning methods / [Deep RL algorithms and representations](https://mlsanity.com/deeprl/deep-rl-algorithms-and-representations/)

## Builds on (most cited first, 30 of 35)

- [Layer Normalization](https://mlsanity.com/paper/1607.06450/) (2016)
- [Continuous control with deep reinforcement learning](https://mlsanity.com/paper/1509.02971/) (2015)
- [Addressing Function Approximation Error in Actor-Critic Methods](https://mlsanity.com/paper/1802.09477/) (2018)
- [High-Dimensional Continuous Control Using Generalized Advantage Estimation](https://mlsanity.com/paper/1506.02438/) (2015)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [Soft Actor-Critic Algorithms and Applications](https://mlsanity.com/paper/1812.05905/) (2018)
- [Deep Reinforcement Learning at the Edge of the Statistical Precipice](https://mlsanity.com/paper/2108.13264/) (2021)
- [When to Trust Your Model: Model-Based Policy Optimization](https://mlsanity.com/paper/1906.08253/) (2019)
- [CURL: Contrastive Unsupervised Representations for Reinforcement Learning](https://mlsanity.com/paper/2004.04136/) (2020)
- [Reinforcement Learning with Augmented Data](https://mlsanity.com/paper/2004.14990/) (2020)
- [Temporal Difference Learning for Model Predictive Control](https://mlsanity.com/paper/2203.04955/) (2022)
- [dm_control: Software and Tasks for Continuous Control](https://mlsanity.com/paper/2006.12983/) (2020)
- [Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning](https://mlsanity.com/paper/2107.09645/) (2021)
- [Efficient Online Reinforcement Learning with Offline Data](https://mlsanity.com/paper/2302.02948/) (2023)
- [The Primacy Bias in Deep Reinforcement Learning](https://mlsanity.com/paper/2205.07802/) (2022)
- [Sample-Efficient Reinforcement Learning with Stochastic Ensemble Value Expansion](https://mlsanity.com/paper/1807.01675/) (2018)
- [Continuous Deep Q-Learning with Model-based Acceleration](https://mlsanity.com/paper/1603.00748/) (2016)
- [EPOpt: Learning Robust Neural Network Policies Using Model Ensembles](https://mlsanity.com/paper/1610.01283/) (2016)
- [Understanding and Preventing Capacity Loss in Reinforcement Learning](https://mlsanity.com/paper/2204.09560/) (2022)
- [Accelerated Policy Learning with Parallel Differentiable Simulation](https://mlsanity.com/paper/2204.07137/) (2022)
- [A Survey on Model-based Reinforcement Learning](https://mlsanity.com/paper/2206.09328/) (2022)
- [The Value-Improvement Path: Towards Better Representations for Reinforcement Learning](https://mlsanity.com/paper/2006.02243/) (2020)
- [Learning Off-Policy with Online Planning](https://mlsanity.com/paper/2008.10066/) (2020)
- [Overestimation, Overfitting, and Plasticity in Actor-Critic: the Bitter Lesson of Reinforcement Learning](https://mlsanity.com/paper/2403.00514/) (2024)
- [Learning to Perform Physics Experiments via Deep Reinforcement Learning](https://mlsanity.com/paper/1611.01843/) (2016)
- [On the model-based stochastic value gradient for continuous reinforcement learning](https://mlsanity.com/paper/2008.12775/) (2020)
- [Bidirectional Model-based Policy Optimization](https://mlsanity.com/paper/2007.01995/) (2020)
- [Simplifying Model-based RL: Learning Representations, Latent-space Models, and Policies with One Objective](https://mlsanity.com/paper/2209.08466/) (2022)
- [Is Model Ensemble Necessary? Model-based RL via a Single Model with Lipschitz Regularized Value Function](https://mlsanity.com/paper/2302.01244/) (2023)
- [Live in the Moment: Learning Dynamics Model Adapted to Evolving Policy](https://mlsanity.com/paper/2207.12141/) (2022)

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

- [A Forensic Analysis of Synthetic Data in RL: Diagnosing and Solving Algorithmic Failures in Model-Based Policy Optimization](https://mlsanity.com/paper/2510.01457/) (2025)
- [A Pontryagin Method of Model-based Reinforcement Learning via Hamiltonian Actor-Critic](https://mlsanity.com/paper/2603.28971/) (2026)
- [Soft MPCritic: Value-Based Reinforcement Learning for Non-Differentiable Predictive Control](https://mlsanity.com/paper/2604.01477/) (2026)
- [Physics-enhanced reinforcement learning for real-time optimal control of dynamical systems](https://mlsanity.com/paper/2607.16177/) (2026)

## Abstract (arXiv)

Dyna-style off-policy model-based reinforcement learning (DMBRL) algorithms are a family of techniques for generating synthetic state transition data and thereby enhancing the sample efficiency of off-policy RL algorithms. This paper identifies and investigates a surprising performance gap observed when applying DMBRL algorithms across different benchmark environments with proprioceptive observations. We show that, while DMBRL algorithms perform well in OpenAI Gym, their performance can drop significantly in DeepMind Control Suite (DMC), even though these settings offer similar tasks and identical physics backends. Modern techniques designed to address several key issues that arise in these settings do not provide a consistent improvement across all environments, and overall our results show that adding synthetic rollouts to the training process -- the backbone of Dyna-style algorithms -- significantly degrades performance across most DMC environments. Our findings contribute to a deeper understanding of several fundamental challenges in model-based RL and show that, like many optimization fields, there is no free lunch when evaluating performance across diverse benchmarks in RL.