Skip to content

Dataset · ianncity

1M reasoning traces from KIMI-K2.5

ianncity/KIMI-K2.5-1000000x

One million reasoning chains distilled from KIMI-K2.5, broken down by topic.

What's inside

A set of 1,000,000 reasoning traces distilled from KIMI-K2.5 at a high reasoning level. Subsets cover code (50%), science (20%), mathematics (15%), computer science (5%), logical questions (5%), and creative writing (5%). The card indicates additional completions in PHD-Science and kimiMath200k.jsonl; the language is English. The format is jsonl files of subsets as specified in the card.

What you can do with it

  • SFT model on reasoning traces
  • Fine-tune LoRA for code and math
  • Verify reasoning on scientific questions
Tasks
Text generation, Question answering
Languages
English
Size
100K–1M rows
Downloads this month
857
Likes
275

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ianncity/KIMI-K2.5-1000000x")