1 个仓库
Awesome GitHub RepositoriesCode-Based Problem Solving Assessments
Frameworks for evaluating the ability of AI models to resolve software bugs by generating functional code patches.
Distinct from Geometric Problem Solving: Candidates are focused on human interview techniques or specific mathematical algorithms, not AI-driven software engineering evaluation.
Explore 1 awesome GitHub repository matching artificial intelligence & ml · Code-Based Problem Solving Assessments. Refine with filters or upvote what's useful.
Awesome Code-Based Problem Solving Assessments GitHub Repositories
用 AI 发现最棒的仓库。我们将通过 AI 为您搜索最匹配的仓库。
- princeton-nlp/swe-bench
princeton-nlp/SWE-bench
5,263在 GitHub 上查看SWE-bench 是一个软件工程基准测试和评估框架，旨在衡量大语言模型解决现实 GitHub Issue 的能力。它提供数据集和评估套件，以验证模型生成的代码补丁是否正确修复了软件 Bug。 该项目包含一个多模态基准测试，用于测试视觉语言模型处理涉及图形界面的 Issue。它利用一系列预处理的仓库 Issue 和黄金标准补丁来训练和测试 AI 编码智能体。 该框架为容器化补丁验证和可重复执行环境提供了基础设施，以确保不同系统间测试的一致性。它涵盖了自动化测试用例执行、补丁预测验证，以及从现实 Bug 报告中生成监督学习数据集。
A process to assess how language models resolve software issues by generating code patches based on a provided codebase.
Python
