Dataset · SynthLabsAI
250k verified math problems
SynthLabsAI/Big-Math-RL-Verified
Large dataset of math problems filtered for RL training of language models.
What's inside
Big-Math is an open dataset of math problems collected for reinforcement learning in language models. The card specifies over 250k strictly filtered and verified problems; the language is English. The data is intended for question-answering and text-generation tasks.
What you can do with it
- RL training of math reasoning models
- SFT and task solver evaluation
- Benchmark for LLMs on mathematics
- Tasks
- Question answering, Text generation
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 8.6K
- Likes
- 245
- Access
- on request on Hugging Face
Models trained on it
- mradermacher/TutorRL-7B-i1-GGUF 1.5K
- eth-nlped/TutorRL-7BText generation 1.1K
- mradermacher/TutorRL-7B-think-i1-GGUF 877
- mradermacher/TutorRL-7B-GGUF 874
- mradermacher/TutorRL-7B-think-GGUF 695
- mradermacher/Raptor-X2-GGUF 629
- viai957/nano-mistral-o1-grpo-stageb-step48Text generation 358
- amd/Instella-3B-MathText generation 105
How to load it
from datasets import load_dataset
ds = load_dataset("SynthLabsAI/Big-Math-RL-Verified")