Siavash Golkar

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.

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.

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.

Christian Pedersen, Tiberiu Tesileanu, Tinghui Wu et al. · cambridge

In Elmarakeby et al., "Biologically informed deep neural network for prostate cancer discovery", a feedforward neural network with biologically informed, sparse connections (P-NET) was presented to model the state of prostate cancer. We verified the reproducibility of the study conducted by Elmarakeby et al., using both their original codebase, and our own re-implementation using more up-to-date libraries. We quantified the contribution of network sparsification by Reactome biological pathways, and confirmed its importance to P-NET's superior performance. Furthermore, we explored alternative neural architectures and approaches to incorporating biological information into the networks. We experimented with three types of graph neural networks on the same training data, and investigated the clinical prediction agreement between different models. Our analyses demonstrated that deep neural networks with distinct architectures make incorrect predictions for individual patient that are persistent across different initializations of a specific neural architecture. This suggests that different neural architectures are sensitive to different aspects of the data, an important yet under-explored challenge for clinical prediction tasks.

Siavash Golkar, David Lipshutz, Tiberiu Tesileanu et al.

Finding informative low-dimensional representations that can be computed efficiently in large datasets is an important problem in data analysis. Recently, contrastive Principal Component Analysis (cPCA) was proposed as a more informative generalization of PCA that takes advantage of contrastive learning. However, the performance of cPCA is sensitive to hyper-parameter choice and there is currently no online algorithm for implementing cPCA. Here, we introduce a modified cPCA method, which we denote cPCA*, that is more interpretable and less sensitive to the choice of hyper-parameter. We derive an online algorithm for cPCA* and show that it maps onto a neural network with local learning rules, so it can potentially be implemented in energy efficient neuromorphic hardware. We evaluate the performance of our online algorithm on real datasets and highlight the differences and similarities with the original formulation.

Jacopo Teneggi, S. M. Bargeen A. Turzo, Tanya Marwah et al.

Large language models (LLMs) are capable of emulating reasoning and using tools, creating opportunities for autonomous agents that execute complex scientific tasks. Protein design provides a natural testbed: although machine learning (ML) methods achieve strong results, these are largely restricted to canonical amino acids and narrow objectives, leaving unfilled need for a generalist tool for broad design pipelines. We introduce Agent Rosetta, an LLM agent paired with a structured environment for operating Rosetta, the leading physics-based heteropolymer design software, capable of modeling non-canonical building blocks and geometries. Agent Rosetta iteratively refines designs to achieve user-defined objectives, combining LLM reasoning with Rosetta's generality. We evaluate Agent Rosetta on design with canonical amino acids, matching specialized models and expert baselines, and with non-canonical residues -- where ML approaches fail -- achieving comparable performance. Critically, prompt engineering alone often fails to generate Rosetta actions, demonstrating that environment design is essential for integrating LLM agents with specialized software. Our results show that properly designed environments enable LLM agents to make scientific software accessible while matching specialized tools and human experts.

Abhipsha Das, Nicholas Lourie, Siavash Golkar et al.

The scientific literature's exponential growth makes it increasingly challenging to navigate and synthesize knowledge across disciplines. Large language models (LLMs) are powerful tools for understanding scientific text, but they fail to capture detailed relationships across large bodies of work. Unstructured approaches, like retrieval augmented generation, can sift through such corpora to recall relevant facts; however, when millions of facts influence the answer, unstructured approaches become cost prohibitive. Structured representations offer a natural complement -- enabling systematic analysis across the whole corpus. Recent work enhances LLMs with unstructured or semistructured representations of scientific concepts; to complement this, we try extracting structured representations using LLMs. By combining LLMs' semantic understanding with a schema of scientific concepts, we prototype a system that answers precise questions about the literature as a whole. Our schema applies across scientific fields and we extract concepts from it using only 20 manually annotated abstracts. To demonstrate the system, we extract concepts from 30,000 papers on arXiv spanning astrophysics, fluid dynamics, and evolutionary biology. The resulting database highlights emerging trends and, by visualizing the knowledge graph, offers new ways to explore the ever-growing landscape of scientific knowledge. Demo: abby101/surveyor-0 on HF Spaces. Code: https://github.com/chiral-carbon/kg-for-science.

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.

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.

Rudy Morel, Francesco Pio Ramunno, Jeff Shen et al.

Conditional diffusion models provide a natural framework for probabilistic prediction of dynamical systems and have been successfully applied to fluid dynamics and weather prediction. However, in many settings, the available information at a given time represents only a small fraction of what is needed to predict future states, either due to measurement uncertainty or because only a small fraction of the state can be observed. This is true for example in solar physics, where we can observe the Sun's surface and atmosphere, but its evolution is driven by internal processes for which we lack direct measurements. In this paper, we tackle the probabilistic prediction of partially observable, long-memory dynamical systems, with applications to solar dynamics and the evolution of active regions. We show that standard inference schemes, such as autoregressive rollouts, fail to capture long-range dependencies in the data, largely because they do not integrate past information effectively. To overcome this, we propose a multiscale inference scheme for diffusion models, tailored to physical processes. Our method generates trajectories that are temporally fine-grained near the present and coarser as we move farther away, which enables capturing long-range temporal dependencies without increasing computational cost. When integrated into a diffusion model, we show that our inference scheme significantly reduces the bias of the predicted distributions and improves rollout stability.

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.

Sofiane Ennadir, Siavash Golkar, Leopoldo Sarra

Self-supervised learning has seen great success recently in unsupervised representation learning, enabling breakthroughs in natural language and image processing. However, these methods often rely on autoregressive and masked modeling, which aim to reproduce masked information in the input, which can be vulnerable to the presence of noise or confounding variables. To address this problem, Joint-Embedding Predictive Architectures (JEPA) has been introduced with the aim to perform self-supervised learning in the latent space. To leverage these advancements in the domain of time series, we introduce Time Series JEPA (TS-JEPA), an architecture specifically adapted for time series representation learning. We validate TS-JEPA on both classification and forecasting, showing that it can match or surpass current state-of-the-art baselines on different standard datasets. Notably, our approach demonstrates a strong performance balance across diverse tasks, indicating its potential as a robust foundation for learning general representations. Thus, this work lays the groundwork for developing future time series foundation models based on Joint Embedding.

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.

Johannes Friedrich, Siavash Golkar, Shiva Farashahi et al.

A major problem in motor control is understanding how the brain plans and executes proper movements in the face of delayed and noisy stimuli. A prominent framework for addressing such control problems is Optimal Feedback Control (OFC). OFC generates control actions that optimize behaviorally relevant criteria by integrating noisy sensory stimuli and the predictions of an internal model using the Kalman filter or its extensions. However, a satisfactory neural model of Kalman filtering and control is lacking because existing proposals have the following limitations: not considering the delay of sensory feedback, training in alternating phases, and requiring knowledge of the noise covariance matrices, as well as that of systems dynamics. Moreover, the majority of these studies considered Kalman filtering in isolation, and not jointly with control. To address these shortcomings, we introduce a novel online algorithm which combines adaptive Kalman filtering with a model free control approach (i.e., policy gradient algorithm). We implement this algorithm in a biologically plausible neural network with local synaptic plasticity rules. This network performs system identification and Kalman filtering, without the need for multiple phases with distinct update rules or the knowledge of the noise covariances. It can perform state estimation with delayed sensory feedback, with the help of an internal model. It learns the control policy without requiring any knowledge of the dynamics, thus avoiding the need for weight transport. In this way, our implementation of OFC solves the credit assignment problem needed to produce the appropriate sensory-motor control in the presence of stimulus delay.

Siavash Golkar, David Lipshutz, Yanis Bahroun et al.

The backpropagation algorithm is an invaluable tool for training artificial neural networks; however, because of a weight sharing requirement, it does not provide a plausible model of brain function. Here, in the context of a two-layer network, we derive an algorithm for training a neural network which avoids this problem by not requiring explicit error computation and backpropagation. Furthermore, our algorithm maps onto a neural network that bears a remarkable resemblance to the connectivity structure and learning rules of the cortex. We find that our algorithm empirically performs comparably to backprop on a number of datasets.

Siavash Golkar, David Lipshutz, Yanis Bahroun et al.

To guide behavior, the brain extracts relevant features from high-dimensional data streamed by sensory organs. Neuroscience experiments demonstrate that the processing of sensory inputs by cortical neurons is modulated by instructive signals which provide context and task-relevant information. Here, adopting a normative approach, we model these instructive signals as supervisory inputs guiding the projection of the feedforward data. Mathematically, we start with a family of Reduced-Rank Regression (RRR) objective functions which include Reduced Rank (minimum) Mean Square Error (RRMSE) and Canonical Correlation Analysis (CCA), and derive novel offline and online optimization algorithms, which we call Bio-RRR. The online algorithms can be implemented by neural networks whose synaptic learning rules resemble calcium plateau potential dependent plasticity observed in the cortex. We detail how, in our model, the calcium plateau potential can be interpreted as a backpropagating error signal. We demonstrate that, despite relying exclusively on biologically plausible local learning rules, our algorithms perform competitively with existing implementations of RRMSE and CCA.

David Lipshutz, Charlie Windolf, Siavash Golkar et al.

Learning latent features from time series data is an important problem in both machine learning and brain function. One approach, called Slow Feature Analysis (SFA), leverages the slowness of many salient features relative to the rapidly varying input signals. Furthermore, when trained on naturalistic stimuli, SFA reproduces interesting properties of cells in the primary visual cortex and hippocampus, suggesting that the brain uses temporal slowness as a computational principle for learning latent features. However, despite the potential relevance of SFA for modeling brain function, there is currently no SFA algorithm with a biologically plausible neural network implementation, by which we mean an algorithm operates in the online setting and can be mapped onto a neural network with local synaptic updates. In this work, starting from an SFA objective, we derive an SFA algorithm, called Bio-SFA, with a biologically plausible neural network implementation. We validate Bio-SFA on naturalistic stimuli.

David Lipshutz, Yanis Bahroun, Siavash Golkar et al.

Cortical pyramidal neurons receive inputs from multiple distinct neural populations and integrate these inputs in separate dendritic compartments. We explore the possibility that cortical microcircuits implement Canonical Correlation Analysis (CCA), an unsupervised learning method that projects the inputs onto a common subspace so as to maximize the correlations between the projections. To this end, we seek a multi-channel CCA algorithm that can be implemented in a biologically plausible neural network. For biological plausibility, we require that the network operates in the online setting and its synaptic update rules are local. Starting from a novel CCA objective function, we derive an online optimization algorithm whose optimization steps can be implemented in a single-layer neural network with multi-compartmental neurons and local non-Hebbian learning rules. We also derive an extension of our online CCA algorithm with adaptive output rank and output whitening. Interestingly, the extension maps onto a neural network whose neural architecture and synaptic updates resemble neural circuitry and synaptic plasticity observed experimentally in cortical pyramidal neurons.

Siavash Golkar

Motivated by the flexibility of biological neural networks whose connectivity structure changes significantly during their lifetime, we introduce the Unstructured Recursive Network (URN) and demonstrate that it can exhibit similar flexibility during training via gradient descent. We show empirically that many of the different neural network structures commonly used in practice today (including fully connected, locally connected and residual networks of different depths and widths) can emerge dynamically from the same URN. These different structures can be derived using gradient descent on a single general loss function where the structure of the data and the relative strengths of various regulator terms determine the structure of the emergent network. We show that this loss function and the regulators arise naturally when considering the symmetries of the network as well as the geometric properties of the input data.

Siavash Golkar, Kyunghyun Cho

We introduce a novel algorithm for the detection of possible sample corruption such as mislabeled samples in a training dataset given a small clean validation set. We use a set of inclusion variables which determine whether or not any element of the noisy training set should be included in the training of a network. We compute these inclusion variables by optimizing the performance of the network on the clean validation set via "gradient descent on gradient descent" based learning. The inclusion variables as well as the network trained in such a way form the basis of our methods, which we call Corruption Detection via Gradient Descent (CDGD). This algorithm can be applied to any supervised machine learning task and is not limited to classification problems. We provide a quantitative comparison of these methods on synthetic and real world datasets.

Kyle Cranmer, Siavash Golkar, Duccio Pappadopulo

We introduce two methods for estimating the density matrix for a quantum system: Quantum Maximum Likelihood and Quantum Variational Inference. In these methods, we construct a variational family to model the density matrix of a mixed quantum state. We also introduce quantum flows, the quantum analog of normalizing flows, which can be used to increase the expressivity of this variational family. The eigenstates and eigenvalues of interest are then derived by optimizing an appropriate loss function. The approach is qualitatively different than traditional lattice techniques that rely on the time dependence of correlation functions that summarize the lattice configurations. The resulting estimate of the density matrix can then be used to evaluate the expectation of an arbitrary operator, which opens the door to new possibilities.

Siavash Golkar, Michael Kagan, Kyunghyun Cho

We introduce Continual Learning via Neural Pruning (CLNP), a new method aimed at lifelong learning in fixed capacity models based on neuronal model sparsification. In this method, subsequent tasks are trained using the inactive neurons and filters of the sparsified network and cause zero deterioration to the performance of previous tasks. In order to deal with the possible compromise between model sparsity and performance, we formalize and incorporate the concept of graceful forgetting: the idea that it is preferable to suffer a small amount of forgetting in a controlled manner if it helps regain network capacity and prevents uncontrolled loss of performance during the training of future tasks. CLNP also provides simple continual learning diagnostic tools in terms of the number of free neurons left for the training of future tasks as well as the number of neurons that are being reused. In particular, we see in experiments that CLNP verifies and automatically takes advantage of the fact that the features of earlier layers are more transferable. We show empirically that CLNP leads to significantly improved results over current weight elasticity based methods.

Siavash Golkar, Kyle Cranmer

We introduce backdrop, a flexible and simple-to-implement method, intuitively described as dropout acting only along the backpropagation pipeline. Backdrop is implemented via one or more masking layers which are inserted at specific points along the network. Each backdrop masking layer acts as the identity in the forward pass, but randomly masks parts of the backward gradient propagation. Intuitively, inserting a backdrop layer after any convolutional layer leads to stochastic gradients corresponding to features of that scale. Therefore, backdrop is well suited for problems in which the data have a multi-scale, hierarchical structure. Backdrop can also be applied to problems with non-decomposable loss functions where standard SGD methods are not well suited. We perform a number of experiments and demonstrate that backdrop leads to significant improvements in generalization.