CPRet-Embeddings

# CPRet-Embeddings This repository provides the **problem descriptions** and their corresponding **precomputed embeddings** used by the [CPRet](https://github.com/coldchair/CPRet) retrieval server. You can explore the retrieval server via the online demo at https://cpret.online/. ## 📦 Files * `probs_2603.jsonl` A JSONL file containing natural language descriptions of competitive programming problems. Each line is a JSON object with metadata such as problem title, platform/source OJ, URL, and full description. * `probs_2603_embs.npy` A NumPy array of dense embeddings corresponding to each problem in `probs_2603.jsonl`. These embeddings are generated using the model [coldchair16/CPRetriever-Prob-Qwen3-4B-2510](https://huggingface.co/coldchair16/CPRetriever-Prob-Qwen3-4B-2510). ## 🧠 Embedding Model The embeddings in this repository are computed with: * [coldchair16/CPRetriever-Prob-Qwen3-4B-2510](https://huggingface.co/coldchair16/CPRetriever-Prob-Qwen3-4B-2510) ## 🌐 Problem Sources (OJ) The current `2603` problem set includes problems from the following online judges: * [Codeforces](https://codeforces.com/) * [Codeforces Gym](https://codeforces.com/gyms) * [AtCoder](https://atcoder.jp/) * [Leetcode](https://leetcode.com/) * [SPOJ](https://www.spoj.com/) * [Nowcoder](https://ac.nowcoder.com/) * [Luogu](https://www.luogu.com.cn/) * [Loj](https://loj.ac/) * [CodeChef](https://www.codechef.com/) * [AIZU](https://judge.u-aizu.ac.jp/onlinejudge/) * [UOJ](https://uoj.ac/) * [QOJ](https://qoj.ac/) Data snapshot: **March 2026**. ## 📚 Full Project For full server usage, deployment instructions, and query examples, see: 👉 [CPRet GitHub Repository](https://github.com/coldchair/CPRet) ## 🗂️ Versioning & Naming Convention This repository uses a **YYMM** naming convention for dataset versions: - `YY` = last two digits of the year - `MM` = month For example, `2603` means **March 2026**. Files are named as: - `probs_<yymm>.jsonl` - `probs_<yymm>_embs.npy` This repository keeps the **latest version** as well as **historical versions** of problem sets and embeddings for reproducibility and backward compatibility. Current latest version: **2603**.# CPRet-嵌入向量（CPRet-Embeddings） 本仓库提供了[CPRet](https://github.com/coldchair/CPRet)检索服务器所使用的**竞赛编程题目描述**及其对应的**预计算嵌入向量**。 您可通过在线演示平台 https://cpret.online/ 体验该检索服务器。 ## 📦 文件列表 * `probs.jsonl` 该JSONL文件存储竞赛编程题目的自然语言描述，文件内每一行均为一个包含题目标题、所属平台、完整题目描述等元数据的JSON对象。 * `probs_embs.npy` 该NumPy数组存储与`probs.jsonl`中各题目对应的稠密嵌入向量，此类嵌入向量由模型[coldchair16/CPRetriever-Prob](https://huggingface.co/coldchair16/CPRetriever-Prob)生成。 ## 📚 完整项目 如需了解该服务器的完整使用方法、部署指南及查询示例，请参阅：👉 [CPRet GitHub仓库](https://github.com/coldchair/CPRet)

搜集汇总

背景与挑战

背景概述

该数据集提供了竞争性编程问题的自然语言描述及其对应的预计算嵌入，用于CPRet检索服务器。它包含JSONL格式的问题描述文件和NumPy格式的嵌入文件，基于特定模型生成，覆盖多个在线判题系统，数据版本为2026年3月。