Dataset · Jackrong
Cleaned reasoning dataset from Kimi-K2.5 model
Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned
Cleaned English and Chinese SFT sample with reasoning chains from KIMI-K2.5.
What's inside
A processed and cleaned version of the KIMI-K2.5-1000000x dataset containing between 100k and 1 million records in English and Chinese. All examples conform to a unified SFT scheme for reasoning training with id, conversations, input, output, domain, and metadata fields, specifying KIMI-K2.5 as the teacher model.
What you can do with it
- SFT fine-tuning of language models for step-by-step reasoning skills
- Distillation of KIMI-K2.5 responses into more compact open LLMs
- Training bilingual (English-Chinese) question-answering assistants
- Tasks
- Text generation, Question answering
- Languages
- English, Chinese
- Size
- 100K–1M rows
- Downloads this month
- 853
- Likes
- 36
Models trained on it
- Jackrong/Qwopus3.6-27B-v1-preview-GGUFImage text → text 7K
- mradermacher/Atem-8B-i1-GGUF 2.6K
- mradermacher/Qwopus3.6-27B-v1-preview-i1-GGUFImage text → text 1.5K
- trjxter/Gwimi-4-12B-IT-GGUFText generation 1.1K
- mradermacher/Atem-v1-1.5B-i1-GGUFText generation 739
- mradermacher/Atem-Pharaoh-3B-i1-GGUFText generation 620
- trjxter/Qwimi-3.6-27B-Coder-MTP-GGUFText generation 570
- mradermacher/Atem-0.6B-i1-GGUF 529
How to load it
from datasets import load_dataset
ds = load_dataset("Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned")