Francois Lanusse

Michael McCabe, Bruno Régaldo-Saint Blancard, Liam Holden Parker et al. · cambridge

We introduce multiple physics pretraining (MPP), an autoregressive task-agnostic pretraining approach for physical surrogate modeling of spatiotemporal systems with transformers. In MPP, rather than training one model on a specific physical system, we train a backbone model to predict the dynamics of multiple heterogeneous physical systems simultaneously in order to learn features that are broadly useful across systems and facilitate transfer. In order to learn effectively in this setting, we introduce a shared embedding and normalization strategy that projects the fields of multiple systems into a shared embedding space. We validate the efficacy of our approach on both pretraining and downstream tasks over a broad fluid mechanics-oriented benchmark. We show that a single MPP-pretrained transformer is able to match or outperform task-specific baselines on all pretraining sub-tasks without the need for finetuning. For downstream tasks, we demonstrate that finetuning MPP-trained models results in more accurate predictions across multiple time-steps on systems with previously unseen physical components or higher dimensional systems compared to training from scratch or finetuning pretrained video foundation models. We open-source our code and model weights trained at multiple scales for reproducibility.

Siavash Golkar, Mariel Pettee, Michael Eickenberg et al. · cambridge

Due in part to their discontinuous and discrete default encodings for numbers, Large Language Models (LLMs) have not yet been commonly used to process numerically-dense scientific datasets. Rendering datasets as text, however, could help aggregate diverse and multi-modal scientific data into a single training corpus, thereby potentially facilitating the development of foundation models for science. In this work, we introduce xVal, a strategy for continuously tokenizing numbers within language models that results in a more appropriate inductive bias for scientific applications. By training specially-modified language models from scratch on a variety of scientific datasets formatted as text, we find that xVal generally outperforms other common numerical tokenization strategies on metrics including out-of-distribution generalization and computational efficiency.

Payel Mukhopadhyay, Stefan S. Nixon, Romain Watteaux et al.

Whether physics foundation models can be usefully deployed on laboratory experiments remains an open question for scientific machine learning (ML). We test this question on the Rayleigh-Taylor instability (RTI), a ubiquitous and demanding fluid instability seen from tabletop flows to supernova explosions, in which small perturbations at a density interface grow into chaotic, multiscale mixing as a lighter fluid accelerates into a heavier one. Standard ML models struggle with RTI, and despite over a century of theoretical, numerical, and experimental work, it carries an unresolved discrepancy between simulation and experiment: the late-time mixing growth rate, $α$, measured in most laboratory experiments ($\sim$ 0.06-0.07), is roughly three times the value from idealized direct numerical simulations (DNS, $\sim$ 0.02). The gap's origin remains debated. These properties make RTI a stringent test for a question that matters well beyond RTI: can foundation models trained only on simulations generalise to sparse, messy, and noisy laboratory settings? We finetune Walrus, a foundation model for continuum dynamics, on three or fewer DNS realizations and recover key RTI physics over long rollouts. Applied zero-shot to sliding-barrier laboratory data, the finetuned model leaves the DNS-like regime and enters the observed growth band, having never seen a single experimental sample. These results provide independent, data-driven evidence that initial conditions play a crucial role in the longstanding sim-experiment gap in $α$. The model also generalises zero-shot to stable stratification, a buoyancy regime absent from training, correctly slowing mixing-layer growth. Together, our results show that foundation models can generalise well beyond their training data, predicting laboratory behavior and unseen physical regimes, opening new ways to probe longstanding simulation-experiment gaps.

Nolan Koblischke, Liam Parker, Francois Lanusse et al.

Finding scientifically interesting phenomena through slow, manual labeling campaigns severely limits our ability to explore the billions of galaxy images produced by telescopes. In this work, we develop a pipeline to create a semantic search engine from completely unlabeled image data. Our method leverages Vision-Language Models (VLMs) to generate descriptions for galaxy images, then contrastively aligns a pre-trained multimodal astronomy foundation model with these embedded descriptions to produce searchable embeddings at scale. We find that current VLMs provide descriptions that are sufficiently informative to train a semantic search model that outperforms direct image similarity search. Our model, AION-Search, achieves state-of-the-art zero-shot performance on finding rare phenomena despite training on randomly selected images with no deliberate curation for rare cases. Furthermore, we introduce a VLM-based re-ranking method that nearly doubles the recall for our most challenging targets in the top-100 results. For the first time, AION-Search enables flexible semantic search scalable to 140 million galaxy images, enabling discovery from previously infeasible searches. More broadly, our work provides an approach for making large, unlabeled scientific image archives semantically searchable, expanding data exploration capabilities in fields from Earth observation to microscopy. The code, data, and app are publicly available at https://github.com/NolanKoblischke/AION-Search

Yesukhei Jagvaral, Francois Lanusse, Rachel Mandelbaum

Forthcoming cosmological imaging surveys, such as the Rubin Observatory LSST, require large-scale simulations encompassing realistic galaxy populations for a variety of scientific applications. Of particular concern is the phenomenon of intrinsic alignments (IA), whereby galaxies orient themselves towards overdensities, potentially introducing significant systematic biases in weak gravitational lensing analyses if they are not properly modeled. Due to computational constraints, simulating the intricate details of galaxy formation and evolution relevant to IA across vast volumes is impractical. As an alternative, we propose a Deep Generative Model trained on the IllustrisTNG-100 simulation to sample 3D galaxy shapes and orientations to accurately reproduce intrinsic alignments along with correlated scalar features. We model the cosmic web as a set of graphs, each graph representing a halo with nodes representing the subhalos/galaxies. The architecture consists of a SO(3) $\times$ $\mathbb{R}^n$ diffusion generative model, for galaxy orientations and $n$ scalars, implemented with E(3) equivariant Graph Neural Networks that explicitly respect the Euclidean symmetries of our Universe. The model is able to learn and predict features such as galaxy orientations that are statistically consistent with the reference simulation. Notably, our model demonstrates the ability to jointly model Euclidean-valued scalars (galaxy sizes, shapes, and colors) along with non-Euclidean valued SO(3) quantities (galaxy orientations) that are governed by highly complex galactic physics at non-linear scales.

Wooseok Ha, Chandan Singh, Francois Lanusse et al.

Recent deep-learning models have achieved impressive prediction performance, but often sacrifice interpretability and computational efficiency. Interpretability is crucial in many disciplines, such as science and medicine, where models must be carefully vetted or where interpretation is the goal itself. Moreover, interpretable models are concise and often yield computational efficiency. Here, we propose adaptive wavelet distillation (AWD), a method which aims to distill information from a trained neural network into a wavelet transform. Specifically, AWD penalizes feature attributions of a neural network in the wavelet domain to learn an effective multi-resolution wavelet transform. The resulting model is highly predictive, concise, computationally efficient, and has properties (such as a multi-scale structure) which make it easy to interpret. In close collaboration with domain experts, we showcase how AWD addresses challenges in two real-world settings: cosmological parameter inference and molecular-partner prediction. In both cases, AWD yields a scientifically interpretable and concise model which gives predictive performance better than state-of-the-art neural networks. Moreover, AWD identifies predictive features that are scientifically meaningful in the context of respective domains. All code and models are released in a full-fledged package available on Github (https://github.com/Yu-Group/adaptive-wavelets).

Siavash Golkar, Jake Kovalic, Irina Espejo Morales et al.

Biological function emerges from coupled constraints across sequence, structure, regulation, evolution, and cellular context, yet most foundation models in biology are trained within one modality or for a fixed forward task. We present MIMIC, a generative multimodal foundation model trained on our newly curated and aligned dataset, LORE, linking nucleic acid, protein, evolutionary, structural, regulatory, and semantic/contextual modalities within partially observed biomolecular states. MIMIC uses a split-track encoder-decoder architecture to condition on arbitrary subsets of observed modalities and reconstruct or generate missing components of molecular state across the genome, transcriptome, and proteome. Multimodal conditioning consistently improves MIMIC's sequence reconstruction relative to sequence-only inputs, while its learned representations enable state-of-the-art performance on RNA and protein downstream tasks. MIMIC achieves state-of-the-art splicing prediction, and its joint generative formulation enables isoform-aware inference that further improves performance. Beyond prediction, the same generative framework supports constrained design. For RNA, MIMIC identifies corrective edits in a clinically relevant HBB splice-disrupting mutation without reverting it by using evolutionary and structural signals. For proteins, jointly conditioning on shape and surface chemistry of PD-L1 and hACE2 binding sites produces diverse, high-confidence sequences with strong in silico support for target binding. Finally, MIMIC uses experimental context as semantic conditioning to model assay-dependent RNA chemical probing, rather than treating context as a fixed output. Together, these results position MIMIC's aligned multimodal generative modeling as a strong foundation for unifying representation learning, conditional prediction, and constrained biomolecular design within a single model.

Yesukhei Jagvaral, Francois Lanusse, Rachel Mandelbaum

Diffusion-based generative models represent the current state-of-the-art for image generation. However, standard diffusion models are based on Euclidean geometry and do not translate directly to manifold-valued data. In this work, we develop extensions of both score-based generative models (SGMs) and Denoising Diffusion Probabilistic Models (DDPMs) to the Lie group of 3D rotations, SO(3). SO(3) is of particular interest in many disciplines such as robotics, biochemistry and astronomy/cosmology science. Contrary to more general Riemannian manifolds, SO(3) admits a tractable solution to heat diffusion, and allows us to implement efficient training of diffusion models. We apply both SO(3) DDPMs and SGMs to synthetic densities on SO(3) and demonstrate state-of-the-art results. Additionally, we demonstrate the practicality of our model on pose estimation tasks and in predicting correlated galaxy orientations for astrophysics/cosmology.

Michael McCabe, Payel Mukhopadhyay, Tanya Marwah et al. · cambridge

Foundation models have transformed machine learning for language and vision, but achieving comparable impact in physical simulation remains a challenge. Data heterogeneity and unstable long-term dynamics inhibit learning from sufficiently diverse dynamics, while varying resolutions and dimensionalities challenge efficient training on modern hardware. Through empirical and theoretical analysis, we incorporate new approaches to mitigate these obstacles, including a harmonic-analysis-based stabilization method, load-balanced distributed 2D and 3D training strategies, and compute-adaptive tokenization. Using these tools, we develop Walrus, a transformer-based foundation model developed primarily for fluid-like continuum dynamics. Walrus is pretrained on nineteen diverse scenarios spanning astrophysics, geoscience, rheology, plasma physics, acoustics, and classical fluids. Experiments show that Walrus outperforms prior foundation models on both short and long term prediction horizons on downstream tasks and across the breadth of pretraining data, while ablation studies confirm the value of our contributions to forecast stability, training throughput, and transfer performance over conventional approaches. Code and weights are released for community use.

Jeff Shen, Francois Lanusse, Liam Holden Parker et al. · cambridge

Sequential scientific data span many resolutions and domains, and unifying them into a common representation is a key step toward developing foundation models for the sciences. Astronomical spectra exemplify this challenge: massive surveys have collected millions of spectra across a wide range of wavelengths and resolutions, yet analyses remain fragmented across spectral domains (e.g., optical vs. infrared) and object types (e.g., stars vs. galaxies), limiting the ability to pool information across datasets. We present a deep learning model that jointly learns from heterogeneous spectra in a self-supervised manner. Our universal spectral tokenizer processes spectra from a variety of object types and resolutions directly on their native wavelength grids, producing intrinsically aligned, homogeneous, and physically meaningful representations that can be efficiently adapted to achieve competitive performance across a range of downstream tasks. For the first time, we demonstrate that a single model can unify spectral data across resolutions and domains, suggesting that our model can serve as a powerful building block for foundation models in astronomy -- and potentially extend to other scientific domains with heterogeneous sequential data, such as climate and healthcare.

Liam Parker, Francois Lanusse, Siavash Golkar et al.

We present AstroCLIP, a single, versatile model that can embed both galaxy images and spectra into a shared, physically meaningful latent space. These embeddings can then be used - without any model fine-tuning - for a variety of downstream tasks including (1) accurate in-modality and cross-modality semantic similarity search, (2) photometric redshift estimation, (3) galaxy property estimation from both images and spectra, and (4) morphology classification. Our approach to implementing AstroCLIP consists of two parts. First, we embed galaxy images and spectra separately by pretraining separate transformer-based image and spectrum encoders in self-supervised settings. We then align the encoders using a contrastive loss. We apply our method to spectra from the Dark Energy Spectroscopic Instrument and images from its corresponding Legacy Imaging Survey. Overall, we find remarkable performance on all downstream tasks, even relative to supervised baselines. For example, for a task like photometric redshift prediction, we find similar performance to a specifically-trained ResNet18, and for additional tasks like physical property estimation (stellar mass, age, metallicity, and sSFR), we beat this supervised baseline by 19\% in terms of $R^2$. We also compare our results to a state-of-the-art self-supervised single-modal model for galaxy images, and find that our approach outperforms this benchmark by roughly a factor of two on photometric redshift estimation and physical property prediction in terms of $R^2$, while remaining roughly in-line in terms of morphology classification. Ultimately, our approach represents the first cross-modal self-supervised model for galaxies, and the first self-supervised transformer-based architectures for galaxy images and spectra.

Benjamin Remy, Francois Lanusse, Niall Jeffrey et al.

Weak lensing mass-mapping is a useful tool to access the full distribution of dark matter on the sky, but because of intrinsic galaxy ellipticies and finite fields/missing data, the recovery of dark matter maps constitutes a challenging ill-posed inverse problem. We introduce a novel methodology allowing for efficient sampling of the high-dimensional Bayesian posterior of the weak lensing mass-mapping problem, and relying on simulations for defining a fully non-Gaussian prior. We aim to demonstrate the accuracy of the method on simulations, and then proceed to applying it to the mass reconstruction of the HST/ACS COSMOS field. The proposed methodology combines elements of Bayesian statistics, analytic theory, and a recent class of Deep Generative Models based on Neural Score Matching. This approach allows us to do the following: 1) Make full use of analytic cosmological theory to constrain the 2pt statistics of the solution. 2) Learn from cosmological simulations any differences between this analytic prior and full simulations. 3) Obtain samples from the full Bayesian posterior of the problem for robust Uncertainty Quantification. We demonstrate the method on the $κ$TNG simulations and find that the posterior mean significantly outperfoms previous methods (Kaiser-Squires, Wiener filter, Sparsity priors) both on root-mean-square error and in terms of the Pearson correlation. We further illustrate the interpretability of the recovered posterior by establishing a close correlation between posterior convergence values and SNR of clusters artificially introduced into a field. Finally, we apply the method to the reconstruction of the HST/ACS COSMOS field and yield the highest quality convergence map of this field to date.

Keming Zhang, Joshua S. Bloom, B. Scott Gaudi et al.

Fast and automated inference of binary-lens, single-source (2L1S) microlensing events with sampling-based Bayesian algorithms (e.g., Markov Chain Monte Carlo; MCMC) is challenged on two fronts: high computational cost of likelihood evaluations with microlensing simulation codes, and a pathological parameter space where the negative-log-likelihood surface can contain a multitude of local minima that are narrow and deep. Analysis of 2L1S events usually involves grid searches over some parameters to locate approximate solutions as a prerequisite to posterior sampling, an expensive process that often requires human-in-the-loop domain expertise. As the next-generation, space-based microlensing survey with the Roman Space Telescope is expected to yield thousands of binary microlensing events, a new fast and automated method is desirable. Here, we present a likelihood-free inference (LFI) approach named amortized neural posterior estimation, where a neural density estimator (NDE) learns a surrogate posterior $\hat{p}(θ|x)$ as an observation-parametrized conditional probability distribution, from pre-computed simulations over the full prior space. Trained on 291,012 simulated Roman-like 2L1S simulations, the NDE produces accurate and precise posteriors within seconds for any observation within the prior support without requiring a domain expert in the loop, thus allowing for real-time and automated inference. We show that the NDE also captures expected posterior degeneracies. The NDE posterior could then be refined into the exact posterior with a downstream MCMC sampler with minimal burn-in steps.

Zaccharie Ramzi, Benjamin Remy, Francois Lanusse et al.

Deep neural networks have proven extremely efficient at solving a wide rangeof inverse problems, but most often the uncertainty on the solution they provideis hard to quantify. In this work, we propose a generic Bayesian framework forsolving inverse problems, in which we limit the use of deep neural networks tolearning a prior distribution on the signals to recover. We adopt recent denoisingscore matching techniques to learn this prior from data, and subsequently use it aspart of an annealed Hamiltonian Monte-Carlo scheme to sample the full posteriorof image inverse problems. We apply this framework to Magnetic ResonanceImage (MRI) reconstruction and illustrate how this approach not only yields highquality reconstructions but can also be used to assess the uncertainty on particularfeatures of a reconstructed image.

Keming Zhang, Joshua S. Bloom, B. Scott Gaudi et al.

Automated inference of binary microlensing events with traditional sampling-based algorithms such as MCMC has been hampered by the slowness of the physical forward model and the pathological likelihood surface. Current analysis of such events requires both expert knowledge and large-scale grid searches to locate the approximate solution as a prerequisite to MCMC posterior sampling. As the next generation, space-based microlensing survey with the Roman Space Observatory is expected to yield thousands of binary microlensing events, a new scalable and automated approach is desired. Here, we present an automated inference method based on neural density estimation (NDE). We show that the NDE trained on simulated Roman data not only produces fast, accurate, and precise posteriors but also captures expected posterior degeneracies. A hybrid NDE-MCMC framework can further be applied to produce the exact posterior.

Chandan Singh, Wooseok Ha, Francois Lanusse et al.

Machine learning lies at the heart of new possibilities for scientific discovery, knowledge generation, and artificial intelligence. Its potential benefits to these fields requires going beyond predictive accuracy and focusing on interpretability. In particular, many scientific problems require interpretations in a domain-specific interpretable feature space (e.g. the frequency domain) whereas attributions to the raw features (e.g. the pixel space) may be unintelligible or even misleading. To address this challenge, we propose TRIM (TRansformation IMportance), a novel approach which attributes importances to features in a transformed space and can be applied post-hoc to a fully trained model. TRIM is motivated by a cosmological parameter estimation problem using deep neural networks (DNNs) on simulated data, but it is generally applicable across domains/models and can be combined with any local interpretation method. In our cosmology example, combining TRIM with contextual decomposition shows promising results for identifying which frequencies a DNN uses, helping cosmologists to understand and validate that the model learns appropriate physical features rather than simulation artifacts.

Francois Lanusse, Peter Melchior, Fred Moolekamp

We present a Bayesian machine learning architecture that combines a physically motivated parametrization and an analytic error model for the likelihood with a deep generative model providing a powerful data-driven prior for complex signals. This combination yields an interpretable and differentiable generative model, allows the incorporation of prior knowledge, and can be utilized for observations with different data quality without having to retrain the deep network. We demonstrate our approach with an example of astronomical source separation in current imaging data, yielding a physical and interpretable model of astronomical scenes.

Siamak Ravanbakhsh, Francois Lanusse, Rachel Mandelbaum et al.

Understanding the nature of dark energy, the mysterious force driving the accelerated expansion of the Universe, is a major challenge of modern cosmology. The next generation of cosmological surveys, specifically designed to address this issue, rely on accurate measurements of the apparent shapes of distant galaxies. However, shape measurement methods suffer from various unavoidable biases and therefore will rely on a precise calibration to meet the accuracy requirements of the science analysis. This calibration process remains an open challenge as it requires large sets of high quality galaxy images. To this end, we study the application of deep conditional generative models in generating realistic galaxy images. In particular we consider variations on conditional variational autoencoder and introduce a new adversarial objective for training of conditional generative networks. Our results suggest a reliable alternative to the acquisition of expensive high quality observations for generating the calibration data needed by the next generation of cosmological surveys.