1/ DSGym: A Holistic Framework for Evaluating and Training Data Science Agents Paper: https://arxiv.org/abs/2601.16344

DSGym is a comprehensive framework designed to evaluate and train AI agents for data science tasks, providing a structured environment for benchmarking agent performance across various data science workflows. The framework enables systematic assessment of how well AI agents can handle end-to-end data science problems, from data exploration to model development. This research from Together AI contributes to advancing autonomous data science capabilities by establishing standardized evaluation metrics and training methodologies.

Source: Together AI (X) | 2026-07-01