# Towards Automated Discovery: A Review of Generative Models, Multimodal Learning and Closed-Loop Workflows in Inverse Materials Design

arXiv: https://arxiv.org/abs/2606.02507 · HTML: https://mlsanity.com/paper/2606.02507/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 89 references that are in the map.
- First posted on arXiv 1 Jun 2026.

## Builds on (most cited first, 24 of 24)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Elucidating the Design Space of Diffusion-Based Generative Models](https://mlsanity.com/paper/2206.00364/) (2022)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift](https://mlsanity.com/paper/1906.02530/) (2019)
- [Automatic chemical design using a data-driven continuous representation of molecules](https://mlsanity.com/paper/1610.02415/) (2016)
- [A Tutorial on Bayesian Optimization](https://mlsanity.com/paper/1807.02811/) (2018)
- [The CMA Evolution Strategy: A Tutorial](https://mlsanity.com/paper/1604.00772/) (2016)
- [Crystal Diffusion Variational Autoencoder for Periodic Material Generation](https://mlsanity.com/paper/2110.06197/) (2021)
- [Crystal Structure Prediction by Joint Equivariant Diffusion](https://mlsanity.com/paper/2309.04475/) (2023)
- [Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks](https://mlsanity.com/paper/1512.07666/) (2015)
- [Predicting materials properties without crystal structure: Deep representation learning from stoichiometry](https://mlsanity.com/paper/1910.00617/) (2019)
- [An invertible crystallographic representation for general inverse design of inorganic crystals with targeted properties](https://mlsanity.com/paper/2005.07609/) (2020)
- [Deep Reinforcement Learning for Inverse Inorganic Materials Design](https://mlsanity.com/paper/2210.11931/) (2022)
- [Multimodal Foundation Models for Material Property Prediction and Discovery](https://mlsanity.com/paper/2312.00111/) (2023)
- [CrysMMNet: Multimodal Representation for Crystal Property Prediction](https://mlsanity.com/paper/2307.05390/) (2023)
- [Bridging Text and Crystal Structures: Literature-driven Contrastive Learning for Materials Science](https://mlsanity.com/paper/2501.12919/) (2025)
- [InvDesFlow-AL: active learning-based workflow for inverse design of functional materials](https://mlsanity.com/paper/2505.09203/) (2025)
- [MEIDNet: Multimodal generative AI framework for inverse materials design](https://mlsanity.com/paper/2601.22009/) (2026)
- [VibroML: an automated toolkit for high-throughput vibrational analysis and dynamic instability remediation of crystalline materials using machine-learned potentials](https://mlsanity.com/paper/2604.27685/) (2026)

## Abstract (arXiv)

Inverse materials design is shifting materials discovery from forward prediction toward targeted proposal of candidates that satisfy objectives under physical constraints. Here, we review advances in generative crystal structure modeling, multimodal learning, and closed-loop design pipelines for crystalline solids. We survey how generators learn chemical-structural priors from databases to enable controllable sampling of periodic structures, comparing variational autoencoders, normalizing flows, autoregressive models, and diffusion models. Across these families, we examine where feasibility constraints and physical priors enter, from representations and training objectives to sampling-time guidance, screening, and relaxation. We also discuss multimodal learning combining crystal structures, thermodynamic and electronic information, microscopy, spectroscopy, processing context, and scientific text to construct materials representations. Inverse-design strategies integrating conditional generation with latent optimization, Bayesian optimization, reinforcement learning, and active learning are also examined. We highlight recurring failure modes, including surrogate exploitation, diversity collapse, distribution shift, and the stability-synthesizability gap, and outline evaluation based on validity, novelty, uniqueness, stability, and cost. To support credible claims, we define a nine-rung discovery-credibility ladder and propose a minimum reporting standard: declared matching tolerances and database snapshots; separate reporting of uniqueness, training-set memorization, and external rediscovery; novelty as a continuous distance distribution; energy-above-hull distributions with functional and hull version; relaxation-survival and dynamical stability rates; and validation cost per credible hit. Headline validity or S.U.N. rates without these disclosures should be treated as uninformative.