Tachibana3-Part1-DeepSeek-V3.1-Terminus

**[Click here to support our open-source dataset and model releases!](https://huggingface.co/spaces/sequelbox/SupportOpenSource)**

**Tachibana3-Part1-DeepSeek-V3.1-Terminus** is a dataset focused on high-difficulty code production tasks, testing the limits of [DeepSeek V3.1 Terminus's](https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus) code-reasoning skills!

This dataset contains 9.3k high-difficulty code-production prompts:
- Questions prioritize real-world, challenging coding tasks across a variety of programming languages and topics.
- Areas of focus include back-end and front-end development, mobile, gamedev, cloud, QA, custom tooling, and embedded systems.
- A wide variety of emphasized languages improves development capability: Python, C, C++, C#, TypeScript, Java, JavaScript, Go, Haskell, R, Ruby, SQL, shell scripts, assembly code, and more!
- Some questions test the model's ability to follow very specific coding instructions, while others provide general business requirements and leave the specific implementation to the model.
- Responses demonstrate the code-reasoning capabilities of DeepSeek's 685b parameter V3.1 Terminus model in reasoning mode.

**Responses have not been edited at all:** the Tachibana dataset strives to accurately represent the V3.1 model. Potential issues may include inaccurate answers and infinite thought loops. Tachibana 3 is presented as-is to be used at your discretion.

**Tachibana 3 is a multi-part dataset;** additional coding queries answered by DeepSeek-V3.2 [can be found here.](https://huggingface.co/datasets/sequelbox/Tachibana3-Part2-DeepSeek-V3.2) The sorting of responses into the two parts is approximate; 1-3% of rows may be placed in the wrong part of the dataset (actually answered by V3.2 instead of V3.1-Terminus.) There is no meaningful impact to the user.

Users should consider applying their own sub-filtering and manual examination of the dataset before use in training.

Do as you will.

**[点击此处支持我们的开源数据集与模型发布！](https://huggingface.co/spaces/sequelbox/SupportOpenSource)**

**Tachibana3-Part1-DeepSeek-V3.1-Terminus** 是一款聚焦高难度代码生成任务的数据集，用于测试[DeepSeek V3.1 Terminus（https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus）]的代码推理能力上限！

本数据集包含9300条高难度代码生成提示词：
- 题目优先涵盖多编程语言与多技术领域的真实场景下的高挑战性编码任务。
- 重点领域涵盖前后端开发、移动开发、游戏开发、云计算、质量保证（QA）、自定义工具开发与嵌入式系统。
- 覆盖丰富编程语言以全面提升开发能力：Python、C、C++、C#、TypeScript、Java、JavaScript、Go、Haskell、R、Ruby、SQL、Shell脚本、汇编代码等！
- 部分题目用于测试模型严格遵循具体编码指令的能力，其余题目仅提供通用业务需求，将具体实现环节交由模型自主完成。
- 测试结果展示了DeepSeek 6850亿参数V3.1 Terminus模型在推理模式下的代码推理性能。

**所有回复均未经过任何编辑**：Tachibana数据集旨在准确呈现V3.1模型的真实表现。测试过程中可能出现答案不准确、无限思维循环等潜在问题。Tachibana 3以原始形态发布，用户可自行决定使用方式。

**Tachibana 3为多部分组成的数据集**；由DeepSeek-V3.2生成的额外编码查询可[在此处获取](https://huggingface.co/datasets/sequelbox/Tachibana3-Part2-DeepSeek-V3.2)。将回复划分为两部分的方式为近似划分，约1%至3%的样本可能被划分至错误的数据集部分（实际由V3.2而非V3.1-Terminus生成），但该情况不会对用户造成实质性影响。

用户在将该数据集用于模型训练前，应考虑自行进行子筛选与人工检查。

尽可随心使用。

搜集汇总

背景与挑战

背景概述

该数据集专注于高难度代码生成任务，包含9.3k个涉及多种编程语言和领域的提示，用于测试DeepSeek V3.1 Terminus模型的代码推理能力。响应未经编辑，可能存在不准确或无限循环问题，且为多部分数据集之一，少量行可能有误分类，但无实质影响。