# Comparing the latent features of universal machine-learning interatomic potentials

arXiv: https://arxiv.org/abs/2512.05717 · HTML: https://mlsanity.com/paper/2512.05717/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 84 references that are in the map.
- First posted on arXiv 5 Dec 2025.

## Builds on (most cited first, 25 of 25)

- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [An Overview of Multi-Task Learning in Deep Neural Networks](https://mlsanity.com/paper/1706.05098/) (2017)
- [Scaling Vision with Sparse Mixture of Experts](https://mlsanity.com/paper/2106.05974/) (2021)
- [The Open Catalyst 2020 (OC20) Dataset and Community Challenges](https://mlsanity.com/paper/2010.09990/) (2020)
- [UMA: A Family of Universal Models for Atoms](https://mlsanity.com/paper/2506.23971/) (2025)
- [Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction](https://mlsanity.com/paper/2502.12147/) (2025)
- [SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials](https://mlsanity.com/paper/2209.10702/) (2022)
- [Unifying machine learning and quantum chemistry -- a deep neural network for molecular wavefunctions](https://mlsanity.com/paper/1906.10033/) (2019)
- [Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models](https://mlsanity.com/paper/2410.12771/) (2024)
- [The Importance of Being Scalable: Improving the Speed and Accuracy of Neural Network Interatomic Potentials Across Chemical Domains](https://mlsanity.com/paper/2410.24169/) (2024)
- [Smooth, exact rotational symmetrization for deep learning on point clouds](https://mlsanity.com/paper/2305.19302/) (2023)
- [A practical guide to machine learning interatomic potentials -- Status and future](https://mlsanity.com/paper/2503.09814/) (2025)
- [Ranking the information content of distance measures](https://mlsanity.com/paper/2104.15079/) (2021)
- [Fast Uncertainty Estimates in Deep Learning Interatomic Potentials](https://mlsanity.com/paper/2211.09866/) (2022)
- [The Open DAC 2023 Dataset and Challenges for Sorbent Discovery in Direct Air Capture](https://mlsanity.com/paper/2311.00341/) (2023)
- [Universally Converging Representations of Matter Across Scientific Foundation Models](https://mlsanity.com/paper/2512.03750/) (2025)
- [Model-free quantification of completeness, uncertainties, and outliers in atomistic machine learning using information theory](https://mlsanity.com/paper/2404.12367/) (2024)
- [Improving Sample and Feature Selection with Principal Covariates Regression](https://mlsanity.com/paper/2012.12253/) (2020)
- [MLIP Arena: Advancing Fairness and Transparency in Machine Learning Interatomic Potentials via an Open, Accessible Benchmark Platform](https://mlsanity.com/paper/2509.20630/) (2025)
- [A prediction rigidity formalism for low-cost uncertainties in trained neural networks](https://mlsanity.com/paper/2403.02251/) (2024)
- [Overcoming systematic softening in universal machine learning interatomic potentials by fine-tuning](https://mlsanity.com/paper/2405.07105/) (2024)
- [The role of feature space in atomistic learning](https://mlsanity.com/paper/2009.02741/) (2020)
- [Uncertainty quantification for predictions of atomistic neural networks](https://mlsanity.com/paper/2207.06916/) (2022)
- [Outlier-Detection for Reactive Machine Learned Potential Energy Surfaces](https://mlsanity.com/paper/2402.17686/) (2024)
- [Learning the electronic density of states in condensed matter](https://mlsanity.com/paper/2006.11803/) (2020)

## Abstract (arXiv)

The past few years have seen the development of ``universal'' machine-learning interatomic potentials (uMLIPs) capable of approximating the ground-state potential energy surface across a wide range of chemical structures and compositions with reasonable accuracy. While these models differ in the architecture and the dataset used, they share the ability to compress a staggering amount of chemical information into descriptive latent features. Herein, we systematically analyze what the different uMLIPs have learned by quantitatively assessing the relative information content of their latent features with feature reconstruction errors, and observing how the trends are affected by the choice of training set and training protocol. We find that uMLIPs encode the chemical space in significantly distinct ways, with substantial cross-model feature reconstruction errors. When variants of the same model architecture are considered, trends become dependent on the dataset, target, and training protocol of choice. We also observe that fine-tuning of a uMLIP retains a strong pre-training bias in the latent features. Finally, we discuss how atom-level features, which are directly output by MLIPs, can be compressed into global structure-level features via concatenation of progressive cumulants, each adding significantly new information about the variability across the atomic environments within a given system.