Jupiter: Enhancing LLM Data Analysis Capabilities via Notebook and Inference-Time Value-Guided Search
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (37) · 31726-31733
DOI: 10.1609/aaai.v40i37.40440 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Shuocheng Li und 9 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Large language models (LLMs) have shown great promise in automating data science workflows. However, existing models still struggle with multi-step reasoning and tool use, limiting their effectiveness on complex data analysis tasks. To address this limitation, we propose a scalable pipeline that extracts high-quality, tool-based data analysis tasks and their executable multi-step solutions from real-world Jupyter notebooks and associated data files. Using this pipeline, we introduce NbQA, a large-scale dataset of standardized task–solution pairs that reflect authentic tool-use patterns in practical data science scenarios. To further enhance the multi-step reasoning capabilities, we present Jupiter, a framework that formulates data analysis as a search problem and applies Monte Carlo Tree Search (MCTS) to generate diverse solution trajectories for value model learning. During inference, Jupiter combines the value model and node visit counts to efficiently collect executable multi-step plans with minimal search steps. Experimental results show that Qwen2.5-7B and 14B-Instruct models on NbQA solve 77.82% and 86.38% of tasks on InfiAgent-DABench, respectively—matching or surpassing GPT-4o and advanced agent frameworks. Further evaluations demonstrate improved generalization and stronger tool-use reasoning across diverse multi-step reasoning tasks.
Zitieren
APA
Li, S., Liu, Y., Du, S., Zeng, W., Xu, Z., Zhou, M., He, Y., Dong, H., Han, S., & Zhang, D. (2026). Jupiter: Enhancing LLM Data Analysis Capabilities via Notebook and Inference-Time Value-Guided Search. Proceedings of the AAAI Conference on Artificial Intelligence, 40(37), 31726-31733. https://doi.org/10.1609/aaai.v40i37.40440
Vancouver
Li S, Liu Y, Du S, Zeng W, Xu Z, Zhou M, et al. Jupiter: Enhancing LLM Data Analysis Capabilities via Notebook and Inference-Time Value-Guided Search. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(37):31726-31733. doi:10.1609/aaai.v40i37.40440
BibTeX
@article{Li2026,
author = {Li, Shuocheng and Liu, Yihao and Du, Silin and Zeng, Wenxuan and Xu, Zhe and Zhou, Mengyu and He, Yeye and Dong, Haoyu and Han, Shi and Zhang, Dongmei},
title = {Jupiter: Enhancing LLM Data Analysis Capabilities via Notebook and Inference-Time Value-Guided Search},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {37},
pages = {31726--31733},
doi = {10.1609/aaai.v40i37.40440}
}
EUVIMED