Seed Dataset
our study four core datasets: the multi-step mathematical word problems of GSM8K, the high-school competition level MATH dataset, and two specialized scientific reasoning datasets in Physics and Chemistry.
Embedded Files
- The GSM8K dataset
We used the train dataset with a total number of 7,413.
Example:
Problem：
Q: Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?\nA:
ground_truth： 72
2. The MATH dataset
2. The MATH dataset
We sampled 2,000 problems from four subsets (Number Theory, Precalculus, Probability, and Algebra) of the MATH dataset.
Example：
problem：
ground_truth： xy
level: Level 4
category: precalculus
3. The Physics dataset
3. The Physics dataset
We collected 1498 physics problems at the high school levels.
Example:
problem:
Calculate the S-wave (secondary wave) velocity in a homogeneous isotropic medium given a shear modulus μ = 3×10^{10} Pa and a density ρ = 2500 kg/m^3
ground_truth: 3.46 × 10^3 m/s
4. The Chemistry dataset
4. The Chemistry dataset
We collected 3,419 chemistry problems at the high school and college levels.
Example:
problem:
What is the net ionic equation for the reaction between hydroiodic acid, and potassium hydroxide?
ground truth: HO- + H+ -> H2O(aq)
Page updated
Google Sites
Report abuse