arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2610.33465cs.AIcs.LG
收藏

LegoSR:智能体引导的自改进科学发现统一框架

LegoSR: A Unified Framework for Self-Improving Agent-Guided Scientific Discovery

The Hong Kong Polytechnic University(香港理工大学);Eastern Institute of Technology(宁波东方理工大学);Amazon Web Services(亚马逊云科技);University of Washington(华盛顿大学);Beijing EEO Education Technology(北京EEO教育科技);Jiangxi University of Finance and Economics(江西财经大学)

Hao Mao, Xu T. Liu, Hao Xu, Peng Zhao, Qian Li, Dongxiao Zhang, Jinyue Yan, Yuntian Chen

AI总结:

提出LegoSR统一框架,集成7种求解器,其智能体可自适应选择求解器,在含噪声微分方程基准中表现优于通用智能体,接近专家调优求解器总和。

摘要:

方程发现求解器通常采用不同的数据接口、方程表示和评估标准,导致其结果难以比较。我们提出LegoSR,这是一个用于代数符号回归和微分方程发现的统一框架。LegoSR通过通用方程表示和评估流程集成了7种求解器,支持受部分已知方程约束的搜索。研究人员可直接运行求解器,或使用LegoSR智能体,该智能体基于求解器知识、中间结果及早期任务经验选择求解器并调整搜索。在含噪声的常微分方程和偏微分方程基准测试中,LegoSR智能体恢复的方程数量接近7种经专家调优求解器的总和,且在相同模型、工具和预算下,比通用智能体的表现高出最多29个百分点。3个示例展示了其基本使用、自适应求解器选择及部分物理知识引导的发现功能。

英文摘要:

Discovering equations from data provides governing laws, constitutive relations and closure models for scientific understanding and prediction. Yet a method that succeeds on one system may fail on another or on noisy data, so discovery depends on expert trial and error. Unsuccessful searches can still reveal useful parts of an equation, but these clues must be recognized and combined. We present LegoSR, a unified framework for agent-guided equation discovery built on composable discovery and cumulative experience. Seven complementary equation discovery methods and shared computational modules form reusable building blocks, connected through a common equation representation. Researchers can launch discovery with one command and guide it with partially known equations. LegoSR Agent selects methods, revises searches using feedback and combines partial findings into new hypotheses. Its Experience Library consolidates graded outcomes into lessons that guide later tasks across equation families. Evaluation covers 134 algebraic regression problems, 40 ordinary differential equation targets and 25 partial differential equation problems at several noise levels, together with three case studies. On noisy PDEs, LegoSR Agent recovers more equations than seven expert-tuned methods combined and stays close to them on ODEs. It also exceeds a generic agent with the same model, tools and budget by up to 28 percentage points, with 63-73% shorter median wall-clock time on PDE problems.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑