Search papers, labs, and topics across Lattice.
This paper introduces DSWorld, a framework designed to enhance the efficiency of autonomous data science agents by predicting the outcomes of data science operations before execution. By employing a Data Science World Model that anticipates state transitions based on current workflows and candidate operations, the authors achieve significant improvements in training and inference speed. Experimental results demonstrate that DSWorld accelerates reinforcement learning-based agent training by approximately 14 times and outperforms the leading LLM baseline by 35.6% on transition prediction tasks.
DSWorld accelerates RL training by 14x while outperforming top LLMs in predicting data science operation outcomes.
Despite strong capabilities in data understanding and decision-making, autonomous data science agents still heavily rely on trial-and-error workflows that involve expensive computation. This bottleneck motivates models that can anticipate the effects of data science operations before real execution. In this paper, we introduce the concept of Data Science World Model, which model the data science execution environment by predicting environment state transitions conditioned on current workflow states and candidate operations. We further propose DSWorld, a practical framework that combines structured state construction, cost-aware routing, lightweight real execution, and an LLM-based simulator for expensive operations. To support training, we construct an 8K-scale transition trajectory dataset and introduce Reflective World Model Optimization, an error-aware reinforcement learning strategy for improving transition prediction. Experiments show that DSWorld accelerates RL-based agent training by approximately $14\times$ and search-based inference by approximately $3$-$6\times$ while maintaining competitive performance, and outperforms the strongest LLM baseline by 35.6% on transition prediction tasks. The code is available at https://anonymous.4open.science/r/DSWorld.
