VeryMath
VeryMath is an open-source platform for AI-assisted mathematical research. We build reusable, verifiable, and collaborative workflows for paper reading, computation, formalization, and research automation. VeryMath 是面向 AI 辅助数学研究的开源平台，致力于构建可复用、可验证、可协作的科研流程，覆盖论文阅读、数学计算、形式化验证与自动化研究。
What VeryMath BuildsVeryMath 在构建什么
VeryMath focuses on practical infrastructure for mathematicians and research groups: not only chat-style assistance, but repository-backed processes where questions, assumptions, computations, failed attempts, reviews, and final writing can all be preserved as durable artifacts. VeryMath 关注数学家和研究团队真正可使用的基础设施：不只是聊天式辅助，而是用仓库承载研究过程，让问题、假设、计算、失败尝试、评审意见和最终写作都成为可追踪的持久成果。
Paper Reading论文阅读
Extract definitions, assumptions, proof structure, citation trails, and open problems from research papers.从数学论文中抽取定义、假设、证明结构、引用脉络和开放问题。
Computational Math计算数学
Turn numerical experiments, optimization routines, and symbolic checks into reproducible project files.将数值实验、优化流程和符号检验整理为可复现的项目文件。
Lean FormalizationLean 形式化
Connect informal arguments with formal proof workflows, theorem statements, and verification-oriented review.连接非形式化论证、定理陈述、形式化证明流程和面向验证的审查。
Evolving Research演化式研究
Support long-running research projects through workstreams, reviewer loops, skill libraries, and final-paper gates.通过工作流、评审循环、技能库和最终论文门控支持长期研究项目。
Repository Directory仓库导览
Explore VeryMath's public repositories by research task, from paper reading and computation to formalization, automated research, writing, and long-running project workspaces. 按研究任务浏览 VeryMath 的公开仓库：从论文阅读、数学计算到形式化验证、自动化研究、论文写作和长期项目工作区。
Structured workflows for reading mathematical papers, extracting proof dependencies, and turning papers into reusable AI4Math skills. 用于阅读数学论文、抽取证明依赖，并将论文内容转化为可复用 AI4Math 技能的结构化流程。
Best for: theorem dependency analysis, source-grounded reading notes, and paper-to-skill extraction. 适合：定理依赖分析、基于原文证据的阅读笔记，以及从论文抽取技能。
Skill packages for mathematical optimization modeling, solver setup, LP/MIP/SOCP workflows, and manifold-constrained optimization. 面向数学优化建模、求解器配置、LP/MIP/SOCP 工作流和流形约束优化的技能包。
Best for: modeling mathematical programs, selecting solvers, adapting examples, and reporting evidence from optimization runs. 适合：建立数学规划模型、选择求解器、改写示例，以及汇报优化运行证据。
Computational workflows for numerical evidence, symbolic structure, finite element reasoning, and mathematical invariant computation. 面向数值证据、符号结构、有限元推理和数学不变量计算的计算工作流。
Best for: turning mathematical objects, equations, data, or paper excerpts into reproducible computational evidence. 适合：将数学对象、方程、数据或论文片段转化为可复现的计算证据。
Lean 4 setup, formalization, proof repair, patch review, and optional Lean-specialist backend workflows for coding agents. 面向编码代理的 Lean 4 环境配置、形式化、证明修复、补丁审查和可选 Lean 专家后端流程。
Best for: inspecting Lean projects, preparing mathlib workspaces, repairing proofs, and validating final patches with evidence. 适合：检查 Lean 项目、准备 mathlib 工作区、修复证明，并用证据验证最终补丁。
Agent-led workflows for iterative mathematical experiments, OpenEvolve runs, evaluation loops, and skill refinement. 由智能体驱动的迭代数学实验、OpenEvolve 运行、评估循环和技能改进流程。
Best for: bounded experiment sessions, metric review, short probes, and next-step iteration. 适合：有边界的实验会话、指标审查、短探测和下一步迭代。
Agentic workflows for mathematical problem discovery, research-loop orchestration, and proof-blueprint review. 面向数学问题发现、研究循环编排和证明蓝图审查的智能体工作流。
Best for: turning broad mathematical intent into structured problems, bounded research runs, and reviewed proof plans. 适合：将宽泛的数学意图转化为结构化问题、有边界的研究运行和经过审查的证明计划。
Source-grounded workflows for mathematical paper drafting, proof-obligation review, and submission-ready LaTeX checks. 基于来源证据的数学论文写作、证明义务审查和投稿级 LaTeX 检查流程。
Best for: converting verified materials into paper text and review artifacts without losing source evidence. 适合：在不丢失来源证据的前提下，将已验证材料转化为论文文本和审查材料。
A SageMath skill that helps agents search SageMath documentation, write runnable SageMath Python code, execute checks, and report verified mathematical conclusions. 帮助智能体检索 SageMath 文档、编写可运行 SageMath Python 代码、执行检查并汇报已验证数学结论的技能。
Best for: algebra, number theory, combinatorics, symbolic computation, and SageMath-backed verification. 适合：代数、数论、组合数学、符号计算，以及基于 SageMath 的验证。
A repository-backed mathematical research workspace for coding agents, designed to make goals, messages, workstreams, reports, reviews, and final papers durable. 一个由仓库承载的数学研究工作区，面向编码代理，让目标、消息、工作流、报告、审查和最终论文都可以持久保存。
Best for: starting a long-running AI-assisted math project and coordinating the other AI4Math skill repositories inside a reviewed workflow. 适合：启动长期 AI 辅助数学研究项目，并在经过审查的流程中协调其他 AI4Math 技能仓库。
Choose a Starting Point选择入口
Read, Compute, Formalize阅读、计算、形式化
Start with the AI4Math skill repository closest to the immediate task: paper reading, optimization, computational mathematics, Lean, SageMath, or paper writing.从最接近当前任务的 AI4Math 技能仓库开始：论文阅读、优化、计算数学、Lean、SageMath 或论文写作。
Run a Research Project运行研究项目
Use co-mathematician when the work needs tracked goals, multiple workstreams, reviewer gates, and a final working paper.当研究需要可追踪目标、多条工作流、评审门控和最终工作论文时，使用 co-mathematician。
Browse Everything浏览全部仓库
Use the VeryMath repository list to see the full set of public repositories and future additions.访问 VeryMath 仓库列表 查看全部公开仓库和未来新增项目。
Research Workflow研究流程
The current VeryMath stack is designed around a human mathematician working with repository-aware agents. The repository keeps the process inspectable and lets research output grow from reviewed intermediate artifacts rather than transient conversation. 当前 VeryMath 技术栈围绕“人类数学家 + 仓库感知智能体”设计。仓库让过程可检查，也让研究成果从经过审查的中间材料中逐步生成，而不是停留在一次性对话里。
Join the Ecosystem加入生态
VeryMath welcomes collaborators interested in AI for mathematics, formal methods, optimization, computational research, academic workflows, and open-source research tools. VeryMath 欢迎关注数学 AI、形式化方法、优化、计算研究、学术工作流和开源研究工具的合作者参与。