Precision therapeutics require multimodal adaptive models that generate personalized treatment recommendations. We introduce TxAgent, an AI agent that leverages multi-step reasoning and real-time biomedical knowledge retrieval across a toolbox of 211 tools to analyze drug interactions, contraindications, and patient-specific treatment strategies. TxAgent evaluates how drugs interact at molecular, pharmacokinetic, and clinical levels, identifies contraindications based on patient comorbidities and concurrent medications, and tailors treatment strategies to individual patient characteristics. It retrieves and synthesizes evidence from multiple biomedical sources, assesses interactions between drugs and patient conditions, and refines treatment recommendations through iterative reasoning. It selects tools based on task objectives and executes structured function calls to solve therapeutic tasks that require clinical reasoning and cross-source validation. The ToolUniverse consolidates 211 tools from trusted sources, including all US FDA-approved drugs since 1939 and validated clinical insights from Open Targets. TxAgent outperforms leading LLMs, tool-use models, and reasoning agents across five new benchmarks: DrugPC, BrandPC, GenericPC, TreatmentPC, and DescriptionPC, covering 3,168 drug reasoning tasks and 456 personalized treatment scenarios. It achieves 92.1% accuracy in open-ended drug reasoning tasks, surpassing GPT-4o and outperforming DeepSeek-R1 (671B) in structured multi-step reasoning. TxAgent generalizes across drug name variants and descriptions. By integrating multi-step inference, real-time knowledge grounding, and tool-assisted decision-making, TxAgent ensures that treatment recommendations align with established clinical guidelines and real-world evidence, reducing the risk of adverse events and improving therapeutic decision-making.

Multi-Source Domain Generalization (DG) is the task of training on multiple source domains and achieving high classification performance on unseen target domains. Recent methods combine robust features from web-scale pretrained backbones with new features learned from source data, and this has dramatically improved benchmark results. However, it remains unclear if DG finetuning methods are becoming better over time, or if improved benchmark performance is simply an artifact of stronger pre-training. Prior studies have shown that perceptual similarity to pre-training data correlates with zero-shot performance, but we find the effect limited in the DG setting. Instead, we posit that having perceptually similar data in pretraining is not enough; and that it is how well these data were learned that determines performance. This leads us to introduce the Alignment Hypothesis, which states that the final DG performance will be high if and only if alignment of image and class label text embeddings is high. Our experiments confirm the Alignment Hypothesis is true, and we use it as an analysis tool of existing DG methods evaluated on DomainBed datasets by splitting evaluation data into In-pretraining (IP) and Out-of-pretraining (OOP). We show that all evaluated DG methods struggle on DomainBed-OOP, while recent methods excel on DomainBed-IP. Put together, our findings highlight the need for DG methods which can generalize beyond pretraining alignment. Domain Generalization (DG) addresses the challenge of enabling AI models to generalize from known domains to unseen ones, a critical task given the inevitable distribution shifts between training and real-world deployment (Saenko et al., 2010). DG pipelines typically consist of three stages: pretraining a model on a large, general dataset; finetuning the model with one or more source domains; and finally evaluating the model on target domains that are distinct from source domains.

Advances in time series models are driving a shift from conventional deep learning methods to pre-trained foundational models. While pre-trained transformers and reprogrammed text-based LLMs report state-of-the-art results, the best-performing architectures vary significantly across tasks, and models often have limited scope, such as focusing only on time series forecasting. Models that unify predictive and generative time series tasks under a single framework remain challenging to achieve. We introduce UniTS, a multi-task time series model that uses task tokenization to express predictive and generative tasks within a single model. UniTS leverages a modified transformer block designed to obtain universal time series representations. This design induces transferability from a heterogeneous, multi-domain pre-training dataset-often with diverse dynamic patterns, sampling rates, and temporal scales-to many downstream datasets, which can also be diverse in task specifications and data domains. Across 38 datasets spanning human activity sensors, healthcare, engineering, and finance domains, UniTS model performs favorably against 12 forecasting models, 20 classification models, 18 anomaly detection models, and 16 imputation models, including repurposed text-based LLMs. UniTS demonstrates effective few-shot and prompt learning capabilities when evaluated on new data domains and tasks. In the conventional single-task setting, UniTS outperforms strong task-specialized time series models. The source code and datasets are available at https://github.com/mims-harvard/UniTS.

Domain generalization (DG) is an important problem that learns a model that can generalize to unseen test domains leveraging one or more source domains, under the assumption of shared label spaces. However, most DG methods assume access to abundant source data in the target label space, a requirement that proves overly stringent for numerous real-world applications, where acquiring the same label space as the target task is prohibitively expensive. For this setting, we tackle the multimodal version of the unsupervised domain generalization (UDG) problem, which uses a large task-agnostic unlabeled source dataset, such as LAION-2B during finetuning. Our framework does not explicitly assume any relationship between the source dataset and target task. Instead, it relies only on the premise that the source dataset can be efficiently searched in a joint vision-language space. For this multimodal UDG setting, we propose a novel method to build a small ($<$100K) subset of the source data in three simple steps: (1) diversified retrieval using label names as queries, (2) rank pseudo-labeling, and (3) clustering to find representative samples. To demonstrate the value of studying the multimodal UDG problem, we compare our results against state-of-the-art source-free DG and zero-shot (ZS) methods on their respective benchmarks and show up to 10% improvement in accuracy on 20 diverse target datasets. Additionally, our multi-stage dataset construction method achieves 3% improvement on average over nearest neighbors retrieval. Code is available: https://github.com/Chris210634/mudg

Recent advances in vision-language models have combined contrastive approaches with generative methods to achieve state-of-the-art (SOTA) on downstream inference tasks like zero-shot image classification. However, a persistent issue of these models for image classification is their out-of-distribution (OOD) generalization capabilities. We first show that when an OOD data point is misclassified, the correct class can be typically found in the Top-K predicted classes. In order to steer the model prediction toward the correct class within the top predicted classes, we propose the Image-Caption Encoding (ICE) method, a straightforward approach that directly enforces consistency between the image-conditioned and caption-conditioned predictions at evaluation time only. Intuitively, we take advantage of unique properties of the generated captions to guide our local search for the correct class label within the Top-K predicted classes. We show that our method can be easily combined with other SOTA methods to enhance Top-1 OOD accuracies by 0.5% on average and up to 3% on challenging datasets. Our code: https://github.com/Chris210634/ice

Transfer learning enables the sharing of common knowledge among models for a variety of downstream tasks, but traditional methods suffer in limited training data settings and produce narrow models incapable of effectively generalizing under distribution shifts. Foundation models have recently demonstrated impressive zero-shot inference capabilities and robustness under distribution shifts. However, zero-shot evaluation for these models has been predominantly confined to benchmarks with simple distribution shifts, limiting our understanding of their effectiveness under the more realistic shifts found in practice. Moreover, common fine-tuning methods for these models have yet to be evaluated against vision models in few-shot scenarios where training data is limited. To address these gaps, we present a new recipe for few-shot fine-tuning of the popular vision-language foundation model CLIP and evaluate its performance on challenging benchmark datasets with realistic distribution shifts from the WILDS collection. Our experimentation demonstrates that, while zero-shot CLIP fails to match performance of trained vision models on more complex benchmarks, few-shot CLIP fine-tuning outperforms its vision-only counterparts in terms of in-distribution and out-of-distribution accuracy at all levels of training data availability. This provides a strong incentive for adoption of foundation models within few-shot learning applications operating with real-world data. Code is available at https://github.com/mit-ll/robust-vision-language-finetuning

Multi-source Domain Generalization (DG) measures a classifier's ability to generalize to new distributions of data it was not trained on, given several training domains. While several multi-source DG methods have been proposed, they incur additional complexity during training by using domain labels. Recent work has shown that a well-tuned Empirical Risk Minimization (ERM) training procedure, that is simply minimizing the empirical risk on the source domains, can outperform most existing DG methods. We identify several key candidate techniques to further improve ERM performance, such as better utilization of training data, model parameter selection, and weight-space regularization. We call the resulting method ERM++, and show it significantly improves the performance of DG on five multi-source datasets by over 5% compared to standard ERM, and beats state-of-the-art despite being less computationally expensive. Additionally, we demonstrate the efficacy of ERM++ on the WILDS-FMOW dataset, a challenging DG benchmark. We hope that ERM++ becomes a strong baseline for future DG research. Code is released at https://github.com/piotr-teterwak/erm_plusplus.

Unsupervised domain adaptation (UDA) enables the transfer of models trained on source domains to unlabeled target domains. However, transferring complex time series models presents challenges due to the dynamic temporal structure variations across domains. This leads to feature shifts in the time and frequency representations. Additionally, the label distributions of tasks in the source and target domains can differ significantly, posing difficulties in addressing label shifts and recognizing labels unique to the target domain. Effectively transferring complex time series models remains a formidable problem. We present Raincoat, the first model for both closed-set and universal domain adaptation on complex time series. Raincoat addresses feature and label shifts by considering both temporal and frequency features, aligning them across domains, and correcting for misalignments to facilitate the detection of private labels. Additionally, Raincoat improves transferability by identifying label shifts in target domains. Our experiments with 5 datasets and 13 state-of-the-art UDA methods demonstrate that Raincoat can improve transfer learning performance by up to 16.33% and can handle both closed-set and universal domain adaptation.

There is extensive interest in metric learning methods for image retrieval. Many metric learning loss functions focus on learning a correct ranking of training samples, but strongly overfit semantically inconsistent labels and require a large amount of data. To address these shortcomings, we propose a new metric learning method, called contextual loss, which optimizes contextual similarity in addition to cosine similarity. Our contextual loss implicitly enforces semantic consistency among neighbors while converging to the correct ranking. We empirically show that the proposed loss is more robust to label noise, and is less prone to overfitting even when a large portion of train data is withheld. Extensive experiments demonstrate that our method achieves a new state-of-the-art across four image retrieval benchmarks and multiple different evaluation settings. Code is available at: https://github.com/Chris210634/metric-learning-using-contextual-similarity

In many domains, including healthcare, biology, and climate science, time series are irregularly sampled with variable time between successive observations and different subsets of variables (sensors) are observed at different time points, even after alignment to start events. These data create multiple challenges for prevailing models that assume fully observed and fixed-length feature representations. To address these challenges, it is essential to understand the relationships between sensors and how they evolve over time. It considers both inter-sensor relationships shared across samples and those unique to each sample that can vary with time, and it adaptively estimates misaligned observations based on nearby observations. Multivariate time series are prevalent in a variety of domains including healthcare, space science, cybersecurity, biology, and finance (Ravuri et al., 2021; Sousa et al., 2020; Sezer et al., 2020; Fawaz et al., 2019; Abanda et al., 2019; Tang et al., 2018). Practical issues often exist in collecting sensor measurements that lead to various types of irregularities caused by missing observations, such as cost saving, sensor failures, external forces in physical scenarios, medical interventions, to name a few (Choi et al., 2020). While temporal machine learning models usually assume fully observable and fixed-size input data, irregularly sampled time series raise considerable challenges. For example, the observations of multiple sensors are not well-aligned; the time intervals among adjacent observations are different across sensors; and different samples have different numbers of observations for different subsets of sensors recorded at different time points.