[论文] DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution

论文概要

研究领域: ML 作者: Nan Jia, Wenxuan Shi, Jingqing Ruan, Zhizhao Zeng, Tianyi Zhai, Zhongda Su, Feng Hong, Zhengyu Chen, Wenhao Yan, Haojin Yang, Siqi Dai, Weipeng Zhang, Ke Zeng, Xunliang Cai 发布时间: 2026-10-22 arXiv: 2610.26924

中文摘要

通用智能体越来越被期望执行专业的分析工作，这需要持续的推理、迭代调查和可靠的证据。训练这些能力需要多样且具有挑战性的任务。然而，直接从原始数据生成任务往往忽略了分析过程中出现的更深层次问题，并且提供的验证证据有限。为解决这些局限，我们提出DataPraxis，一个基于经验证工作区演化的分析任务合成框架。该框架以经过验证的发现和工件不断丰富工作区，然后继承和组合这些工作区以发现更深层次的问题。它从积累的发现及其依赖关系构建任务，在演化过程中和任务组装后验证分析结果。利用该框架，我们构建了DataPraxis-5K数据集，包含5000条经过验证的轨迹，涵盖多种领域和输出格式，用于训练DataPraxis-27B和DataPraxis-35B-A3B。DataPraxis-27B在DABStep上达到55.11分，在JobBench上达到51.9分，与GLM-5.2等前沿模型相当或超越。

原文摘要

General-purpose agents are increasingly expected to perform professional analytics work, which requires sustained reasoning, iterative investigation, and reliable evidence. Training these capabilities requires diverse and challenging tasks. However, generating tasks directly from raw data often overlooks deeper questions that emerge through analysis and provides limited evidence for verification. To address these limitations, we introduce DataPraxis, an analytics task synthesis framework based on verified workspace evolution. The framework progressively enriches workspaces with validated findings and artifacts, then inherits and combines these workspaces to uncover deeper questions. It constructs tasks from accumulated findings and their dependencies, verifying analytical results both duri...

#论文 #arXiv #ML #小凯