Shifts: A Dataset of Real Distributional Shift Across Multiple Large-Scale Tasks

TL;DR

The paper introduces the Shifts Dataset, a large-scale, multi-modal benchmark designed to evaluate uncertainty estimates and robustness to real-world distributional shifts across tabular weather data, machine translation, and vehicle motion prediction. It defines a joint evaluation paradigm and retention-based metrics (R-AUC, F1-AUC, F1@95%, ROC-AUC) to quantify both predictive performance and uncertainty quality under shift, without assuming knowledge of the shift type at deployment. Baselines based on ensembles demonstrate that combining models improves robustness and uncertainty calibration across tasks, with detailed results for each modality. By providing standardized partitions, concrete metrics, and baseline results, the work aims to accelerate development of reliable, safe ML systems in industrial settings and across diverse data modalities.

Abstract

There has been significant research done on developing methods for improving robustness to distributional shift and uncertainty estimation. In contrast, only limited work has examined developing standard datasets and benchmarks for assessing these approaches. Additionally, most work on uncertainty estimation and robustness has developed new techniques based on small-scale regression or image classification tasks. However, many tasks of practical interest have different modalities, such as tabular data, audio, text, or sensor data, which offer significant challenges involving regression and discrete or continuous structured prediction. Thus, given the current state of the field, a standardized large-scale dataset of tasks across a range of modalities affected by distributional shifts is necessary. This will enable researchers to meaningfully evaluate the plethora of recently developed uncertainty quantification methods, as well as assessment criteria and state-of-the-art baselines. In this work, we propose the Shifts Dataset for evaluation of uncertainty estimates and robustness to distributional shift. The dataset, which has been collected from industrial sources and services, is composed of three tasks, with each corresponding to a particular data modality: tabular weather prediction, machine translation, and self-driving car (SDC) vehicle motion prediction. All of these data modalities and tasks are affected by real, "in-the-wild" distributional shifts and pose interesting challenges with respect to uncertainty estimation. In this work we provide a description of the dataset and baseline results for all tasks.

Figures (18)

• Figure 1: Retention curves with CatBoost on eval for the Weather Prediction dataset.
• Figure 2: Retention curves using eGLEU on eval data.
• Figure 3: Retention curves for Vehicle Motion Prediction on full eval data.
• Figure 4: Example error retention curves for the three tasks of the Shifts Dataset.
• Figure 5: Examples of F1-Retention curves for the three tasks of the Shifts Dataset.