Home / Label types / LLM data
Reasoning and chain-of-thought data
Step-by-step solutions written by people, and step-level grading of model reasoning, so your model learns to think in clear, correct steps.
ProblemA shop sells pens at ₹12 each. Riya buys 4 pens and pays with ₹100. How much change does she get?
Step 1: Cost of 4 pens = 4 × 12✓ correct
Step 2: 4 × 12 = 46✗ error found
Corrected: 4 × 12 = 48✓ fixed
Step 3: Change = 100 − 48 = 52✓ correct
Final answer: ₹52✓ correct
WriterSolved
ReviewerSteps graded
QARe-solved
What we label
- Written step-by-step solutions
- Step-level correct or incorrect labels
- Finding and fixing errors in model reasoning
- New problems and prompts written to your spec
- Final answer verification
Where it is used
- Maths and logic fine-tuning
- Reasoning evaluation sets
- Tutoring and education apps
- Reward model training
How we check quality
- Every final answer re-solved independently
- Step grades spot-checked by a second reviewer
- Checks for copied or machine-written solutions
- Difficulty and topic balance reported per batch
Delivery formats
We deliver in the format your training pipeline already uses. Need something else? Send us a sample file and we will match it.
JSONLCSVYour own schema
See reasoning data on your own data
Send a small sample and your guidelines. The pilot is free.