Skip to content

Dataset · nvidia

Chemistry questions for RL training of models

nvidia/Nemotron-RL-litmus-bench-v0.2

Open dataset of 49k chemistry questions with short answers for training and evaluating models using RL.

What's inside

English text dataset from NVIDIA (part of the NeMo Gym library) includes 47,118 training and 2,056 test questions requiring short answers. Tasks are compiled from ChEMBL, ExpansionRx, and Biogen databases, with some utilizing RDKit molecular descriptors.

What you can do with it

  • Training language models in chemical reasoning using RL
  • Benchmarking LLM accuracy in computational chemistry tasks
Tasks
Text generation
Languages
English
Size
10K–100K rows
Downloads this month
1.1K
Likes
37

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-RL-litmus-bench-v0.2")