Proteins are central to biological processes and indispensable for living organisms. Accurate representation of proteins is crucial, especially in drug development. Recent advances have applied machine learning for unsupervised protein representation learning. However, these approaches often focus solely on the amino acid sequence of proteins and lack factual knowledge about proteins and their interactions, thus limiting their performance. In this study, we present GOProteinGNN, a novel architecture that enhances protein language models by integrating protein knowledge graph information during the creation of amino acid level representations. Our approach allows for the integration of information at both the individual amino acid level and the entire protein level, enabling a comprehensive and effective learning process through graph-based learning. By doing so, we can capture complex relationships and dependencies between proteins and their functional annotations, resulting in more robust and contextually enriched protein representations. Unlike previous methods, GOProteinGNN uniquely learns the entire protein knowledge graph during training, which allows it to capture broader relational nuances and dependencies beyond mere triplets as done in previous work. We perform a comprehensive evaluation on several downstream tasks, demonstrating that GOProteinGNN consistently outperforms previous methods, showcasing its effectiveness and establishing it as a state-of-the-art solution for protein representation learning. We discuss the practical integration of GOProteinGNN in a laboratory setting for lipid nanoparticle-based drug delivery, aiming to bypass the blood-brain barrier and discover novel components, with positive results observed in mice.

Despite tremendous recent progress, generative video models still struggle to capture real-world motion, dynamics, and physics. We show that this limitation arises from the conventional pixel reconstruction objective, which biases models toward appearance fidelity at the expense of motion coherence. To address this, we introduce VideoJAM, a novel framework that instills an effective motion prior to video generators, by encouraging the model to learn a joint appearance-motion representation. VideoJAM is composed of two complementary units. During training, we extend the objective to predict both the generated pixels and their corresponding motion from a single learned representation. During inference, we introduce Inner-Guidance, a mechanism that steers the generation toward coherent motion by leveraging the model’s own evolving motion prediction as a dynamic guidance signal. Notably, our framework can be applied to any video model with minimal adaptations, requiring no modifications to the training data or scaling of the model. VideoJAM achieves state-of-the-art performance in motion coherence, surpassing highly competitive proprietary models while also enhancing the perceived visual quality of the generations. These findings emphasize that appearance and motion can be complementary and, when effectively integrated, enhance both the visual quality and the coherence of video generation.

Autoregressive language models accumulate errors due to their fixed, irrevocable left-to-right token generation. To address this, we propose a new sampling method called Resample-Previous-Tokens (RPT). RPT mitigates error accumulation by iteratively revisiting and potentially replacing tokens in a window of previously generated text. Fine-tuning a pretrained 8B parameter model with RPT for only 100B resulted in ~10% relative improvements on reasoning and coding benchmarks compared to the standard sampling.

Diffusion and flow matching models have significantly advanced media generation, yet their design space is well-explored, somewhat limiting further improvements. Concurrently, autoregressive (AR) models, particularly those generating continuous tokens, have emerged as a promising direction for unifying text and media generation, showing improved performance at scale. This paper introduces Transition Matching (TM), a novel discrete-time, continuous-state generative paradigm that unifies and advances both diffusion/flow models and continuous AR generation. TM decomposes complex generation tasks into simpler Markov transitions, allowing for expressive non-deterministic probability transition kernels and arbitrary non-continuous supervision processes, thereby unlocking new flexible design avenues. We explore these choices through three TM variants: (i) Difference Transition Matching (DTM), which generalizes flow matching to discrete-time by directly learning transition probabilities, yielding state-of-the-art image quality and text adherence. (ii) Autoregressive Transition Matching (ARTM) and (iii) Full History Transition Matching (FHTM) are partially and fully causal models, respectively, that generalize continuous AR methods. They achieve continuous causal AR generation quality comparable to non-causal approaches and potentially enable seamless integration with existing AR text generation techniques. Notably, FHTM is the first fully causal model to match or surpass the performance of flow-based methods on text-to-image task in continuous domains. We demonstrate these contributions through a rigorous large-scale comparison of TM variants and relevant baselines, maintaining a fixed architecture, training data, and hyperparameters.

Text-to-Video generation has achieved remarkable progress with the rise of diffusion models. In this work, we introduce Cached Memory-Guided Video Generation (Corgi), aiming to generate multi-scene videos with arbi-trary number of video clips, conditioned on input images and instruction prompts. This is a challenging task, as tra-ditional T2V methods often struggle to maintain the quality of longer videos due to the difficulties in preserving visual context from earlier scenes. We address this by introducing a cached memory mechanism that stores the key frames. Our multi-scene video generation process is explicitly con-ditioned on the cached memories to avoid forgetting the vi-sual appearance of target subjects. Corgi shows significant improvement in multi-scene video generation compared to the prior art, with up to 59.2% in long-term consistency and 7.6% in diversity.

Instruction-based image editing holds immense potential for a variety of applications, as it enables users to perform any editing operation using a natural language instruction. However, current models in this domain often struggle with accurately executing user instructions. We present Emu Edit, a multitask image editing model which sets state-of-the-art results in instruction-based image editing. To develop Emu Edit we train it to multitask across an unprecedented range of tasks, such as region-based editing, free-form editing, and Computer Vision tasks, all of which are formulated as generative tasks. Additionally, to enhance Emu Edit's multitask learning abilities, we provide it with learned task embeddings which guide the generation process towards the correct edit type. Both these elements are essential for Emu Edit's outstanding performance. Furthermore, we show that Emu Edit can generalize to new tasks, such as image inpainting, super-resolution, and compositions of editing tasks, with just a few labeled examples. This capability offers a significant advantage in scenarios where high-quality samples are scarce. Lastly, to facilitate a more rigorous and informed assessment of instructable image editing models, we release a new challenging and versatile benchmark that includes seven different image editing tasks.1

Taming the generation outcome of state of the art Diffusion and Flow-Matching (FM) models without having to re-train a task-specific model unlocks a powerful tool for solving inverse problems, conditional generation, and controlled generation in general. In this work we introduce *D-Flow*, a simple framework for controlling the generation process by differentiating through the flow, optimizing for the source (noise) point. We motivate this framework by our key observation stating that for Diffusion/FM models trained with Gaussian probability paths, differentiating through the generation process projects gradient on the data manifold, implicitly injecting the prior into the optimization process. We validate our framework on linear and non-linear controlled generation problems including: image and audio inverse problems and conditional molecule generation reaching state of the art performance across all.

This paper introduces Bespoke Non-Stationary (BNS) Solvers, a solver distillation approach to improve sample efficiency of Diffusion and Flow models. BNS solvers are based on a family of non-stationary solvers that provably subsumes existing numerical ODE solvers and consequently demonstrate considerable improvement in sample approximation (PSNR) over these baselines. Compared to model distillation, BNS solvers benefit from a tiny parameter space ($<$200 parameters), fast optimization (two orders of magnitude faster), maintain diversity of samples, and in contrast to previous solver distillation approaches nearly close the gap from standard distillation methods such as Progressive Distillation in the low-medium NFE regime. For example, BNS solver achieves 45 PSNR / 1.76 FID using 16 NFE in class-conditional ImageNet-64. We experimented with BNS solvers for conditional image generation, text-to-image generation, and text-2-audio generation showing significant improvement in sample approximation (PSNR) in all.

Dynamic temporal graphs evolve over time, adding and removing nodes and edges between time snapshots. The tasks performed on such graphs are diverse and include detecting temporal trends, finding graph-to-graph similarities, and graph visualization and clustering. For all these tasks, it is necessary to embed the entire graph in a low-dimensional space by using graph-level representations instead of the more common node-level representations. This embedding requires handling the appearance of new nodes over time as well as capturing temporal patterns of the entire graph. Most existing methods perform temporal node embeddings and focus on different methods of aggregating them for a graph-based representation. In this work, we propose an end-to-end architecture that captures both the node embeddings and their influence in a structural context during a specific time period of the graph. We present GraphERT (Graph Embedding Representation using Transformers), a novel approach to temporal graph-level embeddings. Our method pioneers the use of Transformers to seamlessly integrate graph structure learning with temporal analysis. By employing a masked language model on sequences of graph random walks, together with a novel temporal classification task, our model not only comprehends the intricate graph dynamics but also unravels the temporal significance of each node and path. This novel training paradigm empowers GraphERT to capture the essence of both the structural and temporal aspects of graphs, surpassing state-of-the-art approaches across multiple tasks on real-world datasets.

Drug development is a long and costly process consisting of several stages that can take many years to complete. One of the early stage's goals is to optimize a novel chemical compound to be active against a target protein associated with the disease. Often machine learning techniques are used to improve the procedure of discovering and optimizing potential drug candidates. The goal of molecule optimization is, given an input molecule, to produce a new molecule that is chemically similar to the input molecule but with an improved property. We present a novel algorithm that during optimization divides a molecule into two disjoint substructures that we call: the molecule chains and the molecule core. Our approach is inspired by expert design of chemical compounds that employ a fundamental molecular template and add to it chemical functional groups to generate compounds with desired properties. We train a model to generate the molecule chains with the desired properties for optimization, which are then attached to the molecule core to construct a novel molecule with high similarity to the input molecule. This is achieved by selective masking of pairs of input molecules' chains and cores during training. Additionally, we demonstrate the extension of this approach to data-scarce tasks, like targeting a drug to a novel protein. We first evaluate our method on standard molecule optimization tasks such as inhibition against glycogen synthase kinase-3 beta (GSK3β). We then empirically compared the model performance with the state-of-the-art algorithms over 21 novel proteins and show superior performance.

We present MAV3D (Make-A-Video3D), a method for generating three-dimensional dynamic scenes from text descriptions. Our approach uses a 4D dynamic Neural Radiance Field (NeRF), which is optimized for scene appearance, density, and motion consistency by querying a Text-to-Video (T2V) diffusion-based model. The dynamic video output generated from the provided text can be viewed from any camera location and angle, and can be composited into any 3D environment. MAV3D does not require any 3D or 4D data and the T2V model is trained only on Text-Image pairs and unlabeled videos. We demonstrate the effectiveness of our approach using comprehensive quantitative and qualitative experiments and show an improvement over previously established internal baselines. To the best of our knowledge, our method is the first to generate 3D dynamic scenes given a text description. Generated samples can be viewed at make-a-video3d.github.io

The ability to collect a large dataset of human preferences from text-to-image users is usually limited to companies, making such datasets inaccessible to the public. To address this issue, we create a web app that enables text-to-image users to generate images and specify their preferences. Using this web app we build Pick-a-Pic, a large, open dataset of text-to-image prompts and real users’ preferences over generated images. We leverage this dataset to train a CLIP-based scoring function, PickScore, which exhibits superhuman performance on the task of predicting human preferences. Then, we test PickScore’s ability to perform model evaluation and observe that it correlates better with human rankings than other automatic evaluation metrics. Therefore, we recommend using PickScore for evaluating future text-to-image generation models, and using Pick-a-Pic prompts as a more relevant dataset than MS-COCO. Finally, we demonstrate how PickScore can enhance existing text-to-image models via ranking.

Graph neural networks (GNNs), has been widely used for supervised learning tasks in graphs reaching state-of-the-art results. However, little work was dedicated to creating unbiased GNNs, i.e., where the classification is uncorrelated with sensitive attributes, such as race or gender. Some ignore the sensitive attributes or optimize for the criteria of statistical parity for fairness. However, it has been shown that neither approaches ensure fairness, but rather cripple the utility of the prediction task. In this work, we present a GNN framework that allows optimizing representations for the notion of Equalized Odds fairness criteria. The architecture is composed of three components: (1) a GNN classifier predicting the utility class, (2) a sampler learning the distribution of the sensitive attributes of the nodes given their labels. It generates samples fed into a (3) discriminator that discriminates between true and sampled sensitive attributes using a novel ``permutation loss'' function. Using these components, we train a model to neglect information regarding the sensitive attribute only with respect to its label. To the best of our knowledge, we are the first to optimize GNNs for the equalized odds criteria. We evaluate our classifier over several graph datasets and sensitive attributes and show our algorithm reaches state-of-the-art results.

Recent global events have emphasized the importance of accelerating the drug discovery process. A way to deal with the issue is to use machine learning to increase the rate at which drugs are made available to the public. However, chemical labeled data for real-world applications is extremely scarce making traditional approaches less effective. A fruitful course of action for this challenge is to pretrain a model using related tasks with large enough datasets, with the next step being finetuning it for the desired task. This is challenging as creating these datasets requires labeled data or expert knowledge. To aid in solving this pressing issue, we introduce MISU - Molecular Inherent SUpervision, a unique method for pretraining graph neural networks for molecular property prediction. Our method leapfrogs past the need for labeled data or any expert knowledge by introducing three innovative components that utilize inherent properties of molecular graphs to induce information extraction at different scales, from the local neighborhood of an atom to substructures in the entire molecule. Our empirical results for six chemical-property-prediction tasks show that our method reaches state-of-the-art results compared to numerous baselines.

Consumer demand forecasting is of high importance for many e-commerce applications, including supply chain optimization, advertisement placement, and delivery speed optimization. However, reliable time series sales forecasting for e-commerce is difficult, especially during periods with many anomalies, as can often happen during pandemics, abnormal weather, or sports events. Although many time series algorithms have been applied to the task, prediction during anomalies still remains a challenge. In this work, we hypothesize that leveraging external knowledge found in world events can help overcome the challenge of prediction under anomalies. We mine a large repository of 40 years of world events and their textual representations. Further, we present a novel methodology based on transformers to construct an embedding of a day based on the relations of the day's events. Those embeddings are then used to forecast future consumer behavior. We empirically evaluate the methods over a large e-commerce products sales dataset, extracted from eBay, one of the world's largest online marketplaces. We show over numerous categories that our method outperforms state-of-the-art baselines during anomalies.

In e-commerce, the watchlist enables users to track items over time and has emerged as a primary feature, playing an important role in users' shopping journey. Watchlist items typically have multiple attributes whose values may change over time (e.g., price, quantity). Since many users accumulate dozens of items on their watchlist, and since shopping intents change over time, recommending the top watchlist items in a given context can be valuable. In this work, we study the watchlist functionality in e-commerce and introduce a novel watchlist recommendation task. Our goal is to prioritize which watchlist items the user should pay attention to next by predicting the next items the user will click. We cast this task as a specialized sequential recommendation task and discuss its characteristics. Our proposed recommendation model, Trans2D, is built on top of the Transformer architecture, where we further suggest a novel extended attention mechanism (Attention2D) that allows to learn complex item-item, attribute-attribute and item-attribute patterns from sequential-data with multiple item attributes. Using a large-scale watchlist dataset from eBay, we evaluate our proposed model, where we demonstrate its superiority compared to multiple state-of-the-art baselines, many of which are adapted for this task.

SUMMARY: How do nuances of scientists' attention influence what they discover? We pursue an understanding of the influences of patterns of attention on discovery with a case study about confirmations of protein-protein interactions over time. We find that modeling and accounting for attention can help us to recognize and interpret biases in large-scale and widely used databases of confirmed interactions and to better understand missing data and unknowns. Additionally, we present an analysis of how awareness of patterns of attention and use of debiasing techniques can foster earlier discoveries. AVAILABILITY AND IMPLEMENTATION: The data is freely available at https://github.com/urielsinger/PPI-unbias.

In this work, we present a method for node embedding in temporal graphs. We propose an algorithm that learns the evolution of a temporal graph's nodes and edges over time and incorporates this dynamics in a temporal node embedding framework for different graph prediction tasks. We present a joint loss function that creates a temporal embedding of a node by learning to combine its historical temporal embeddings, such that it optimizes per given task (e.g., link prediction). The algorithm is initialized using static node embeddings, which are then aligned over the representations of a node at different time points, and eventually adapted for the given task in a joint optimization. We evaluate the effectiveness of our approach over a variety of temporal graphs for the two fundamental tasks of temporal link prediction and multi-label node classification, comparing to competitive baselines and algorithmic alternatives. Our algorithm shows performance improvements across many of the datasets and baselines and is found particularly effective for graphs that are less cohesive, with a lower clustering coefficient.