Skip to content

Dataset · angrygiraffe

Claude synthetic reasoning for QA

angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k

Text QA examples with synthetic reasoning from Claude, without human verification.

What's inside

A set of text examples for text-generation and question-answering in English. Answers and reasoning are generated by Claude, and this is neither a real chain-of-thought nor its summary, but a synthetic imitation of 'thought'. The data has not undergone human verification; metadata estimates the volume at 10-100k examples.

What you can do with it

  • Fine-tuning LLMs on synthetic reasoning
  • Evaluating QA models on Claude responses
  • LoRA experiments for reasoning
Tasks
Text generation, Question answering
Languages
English
Size
10K–100K rows
Downloads this month
1.9K
Likes
465

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k")