Search papers, labs, and topics across Lattice.
EvoDS is a self-evolving autonomous data science agent that enhances automated data science through skill learning and long-term context management. By implementing an Autonomous Skill Acquisition mechanism and an Adaptive Context Compression strategy, EvoDS effectively synthesizes and reuses skills while managing context as a learned control problem. The agent demonstrates a significant performance improvement, outperforming existing data science agents by an average of 28.9% across multiple benchmarks and eliminating out-of-token failures.
EvoDS not only learns new skills autonomously but also manages context intelligently, leading to a 28.9% performance boost over existing data science agents.
Recent progress in Large Language Model (LLM) agents has enabled promising advances in automated data science. However, existing approaches remain fundamentally limited by their static action sets and lack of principled long-horizon context management, hindering their ability to accumulate reusable experience across tasks and operate reliably in multi-stage, iterative data science pipelines. To address these challenges, we introduce EvoDS, a self-evolving autonomous data science agent that learns to expand its skills and adaptively managing long-term context through agentic reinforcement learning. Specifically, EvoDS introduces two key strategies: (1) Autonomous Skill Acquisition (ASA) mechanism, which enables agents to synthesize, validate, and reuse executable skills; and (2) Adaptive Context Compression (ACC) strategy, which treats context management as a learned control problem rather than passive truncation. These strategies are orchestrated within a two-stage multi-agent training scheme, enabling EvoDS to autonomously improve over time. Theoretically, we prove that EvoDS's hierarchical design reduces tool-selection error, and its optimization objective aligns with an information bottleneck principle, ensuring efficient context use. Empirically, EvoDS outperforms state-of-the-art open-source data science agents by an average of 28.9% across four diverse benchmarks while eliminating out-of-token failures. Our code and data are available at https://github.com/usail-hkust/EvoDS.
