yu0226/CipherBank 已完整同步
CipherBank Benchmark
Benchmark description
CipherBank, a comprehensive benchmark designed to evaluate the reasoning capabilities of LLMs in cryptographic decryption tasks. CipherBank comprises 2,358 meticulously crafted problems, covering 262 unique plaintexts across 5 domains and 14 subdomains, with a focus on privacy-sensitive and real-world scenarios that necessitate encryption. From a cryptographic perspective, CipherBank incorporates 3 major categories of encryption methods, spanning 9 distinct algorithms, ranging from classical ciphers to custom cryptographic techniques.
Model Performance
We evaluate state-of-the-art LLMs on CipherBank, e.g., GPT-4o, DeepSeek-V3, and cutting-edge reasoning-focused models such as o1 and DeepSeek-R1. Our results reveal significant gaps in reasoning abilities not only between general-purpose chat LLMs and reasoning-focused LLMs but also in the performance of current reasoning-focused models when applied to classical cryptographic decryption tasks, highlighting the challenges these models face in understanding and manipulating encrypted data.
| Model | CipherBank Score (%) |
|---|---|
| Qwen2.5-72B-Instruct | 0.55 |
| Llama-3.1-70B-Instruct | 0.38 |
| DeepSeek-V3 | 9.86 |
| GPT-4o-mini-2024-07-18 | 1.00 |
| GPT-4o-2024-08-06 | 8.82 |
| gemini-1.5-pro | 9.54 |
| gemini-2.0-flash-exp | 8.65 |
| Claude-Sonnet-3.5-1022 | 45.14 |
| DeepSeek-R1 | 25.91 |
| gemini-2.0-flash-thinking | 13.49 |
| o1-mini-2024-09-12 | 20.07 |
| o1-2024-12-17 | 40.59 |
Please see paper & website for more information:
Citation
If you find CipherBank useful for your research and applications, please cite using this BibTeX:
@misc{li2025cipherbankexploringboundaryllm,
title={CipherBank: Exploring the Boundary of LLM Reasoning Capabilities through Cryptography Challenges},
author={Yu Li and Qizhi Pei and Mengyuan Sun and Honglin Lin and Chenlin Ming and Xin Gao and Jiang Wu and Conghui He and Lijun Wu},
year={2025},
eprint={2504.19093},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2504.19093},
}
11 个文件
浏览文件数据集版权信息
本数据集的许可证为 Apache License 2.0。如有违反相关条款,请联系 WEHUB,我们将及时处理。 查看许可证
通过 WeHub CLI 下载当前数据集快照。下列命令会固定为当前页面展示的数据版本(如果页面提供版本)。文件字节由本机直连存储下载,浏览器不会签发或保存下载链接。
前置要求
需要 Node.js 18 及以上,以及 npm(或 npx)。
1. 安装 CLI
npm install -g wehub-cli@latest
2. 下载此数据集
wehub datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db
若中断或部分失败,在同一目录重新执行同一命令即可续传。默认会校验 SHA-256。
免全局安装
npx --yes wehub-cli@latest datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db
高级选项
以下为 wehub datasets download 已支持的参数示例:
强制重新下载,不复用已校验的本地文件
wehub datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db --overwrite
仅包含匹配路径
wehub datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db --include "*.jsonl"
排除匹配路径
wehub datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db --exclude "*.md"
提高并发下载数
wehub datasets download ds_ext_3561_a9663f79db --revision d2f8f7cc7ca6a0c1b8aa25898747a9b80ed67729 --output ./ds_ext_3561_a9663f79db --jobs 8
完整帮助:wehub datasets download --help