遇见数据集

crucible

收藏
魔搭社区2026-07-20 更新2026-08-09 收录
官方服务:

资源简介:

# crucible A mixed prompt dataset combining reasoning and creative writing, for model distillation. **Total prompts:** 37,460 ## By task type | task_type | rows | |---|---| | verifiable_math | 10,736 | | Artistic - Creative & Narrative Writing | 7,503 | | no_verification | 3,055 | | prime_rl_code | 2,585 | | ifeval | 1,794 | | code_output_prediction | 1,712 | | complex_json_output | 1,105 | | Finance & Business - Strategic Planning and Analysis | 868 | | reasoning_gym | 835 | | Science & Engineering - Analytical Writing | 826 | | Education & Teaching | 529 | | Artistic - Essay Writing & Topic/Argument Writing | 521 | | Non-Science & Non-Engineering Scholarly Paper Writing | 508 | | Science & Engineering - Technical Research Writing | 495 | | Artistic - Analytical & Critical Writing | 486 | | Advertising & Marketing - Content Writing | 443 | | Ordinary Life - Practical & Real-World Guidance | 431 | | Artistic - Scriptwriting for Media & Performance | 390 | | Ordinary Life - Analytical & Creative Reasoning | 312 | | unscramble_sentence | 293 | | Finance & Business - Proposals & Persuasion | 291 | | pydantic_adherance | 289 | | Law - Legal Guidance & Analysis | 241 | | Finance & Business - Reporting & Performance Review | 209 | | Ordinary Life - Factual QAs | 189 | | Science & Engineering - Engineering Report Writing | 162 | | Advertising & Marketing - Promotional Writing | 120 | | Politics | 119 | | Finance & Business - Operational Communication & Agreements | 105 | | Others | 104 | | Advertising & Marketing - Brand Strategy & Identity | 81 | | Law - Judicial & Procedural Documents | 67 | | ascii_tree_formatting | 25 | | Advertising & Marketing - Customer Review & Feedback | 20 | | Science & Engineering - Patent Writing | 11 | ## By language | lang | rows | |---|---| | en | 27,883 | | de | 1,948 | | fr | 1,947 | | ja | 1,945 | | it | 1,944 | | zh | 1,793 | ## Sources - PrimeIntellect/SYNTHETIC-2-SFT-verified — https://huggingface.co/datasets/PrimeIntellect/SYNTHETIC-2-SFT-verified - m-a-p/DeepWriting-20K — https://huggingface.co/datasets/m-a-p/DeepWriting-20K

提供机构:
maas
创建时间:
2026-06-28
二维码
社区交流群
二维码
科研交流群
商业服务