DeepData/Datasets/Undergraduate STEM
Focus line

Tens of millions of university-level problems across mathematics, physics, chemistry, biology and computer science: the backbone for training general scientific and logical reasoning.

Scale today40M+ problems
Target100M
SubjectsMath · Phys · Chem · Bio · CS
Accuracy98%+ sampled
01Why this data

What makes it worth training on.

Scale that tracks compute

40M+ items today, growing toward 100M, because output scales with compute.

Undergraduate STEM: 40M+ today, 100M target →
Training-ready on arrival

Answers and full step-by-step solutions in clean LaTeX, schema-checked.

100% Markdown/LaTeX · schema-checked JSONL →
Built to your specification

Slice by subject, level, question type and difficulty tier.

Scope → pilot → acceptance → scale →
02Sample

Inspect the data.

Real items from this line, shown exactly as delivered. Switch to JSON for the raw record, or download the public sample file.

Samples24 of 100
MathematicsUndergraduateCalculation

Let and be differentiable functions, and let be a differentiable function with respect to its first variable. Define the function Find the derivative in terms of , its partial derivative with respect to , , , and their derivatives.

To compute efficiently, we treat the integral as a composite function of three independent variables. Define

Then the original function can be written as . By the multivariable chain rule:

Using the Fundamental Theorem of Calculus, we compute the partial derivatives of :

Substituting , , and into these partials and then into the chain rule expression gives:

This elegantly derives the Leibniz integral rule without needing to split the integral at a constant base point.

Download the public sample↓ stem-sample.jsonl205 KB
03Coverage

How it is sliced.

Subjects

Mathematics (calculus, linear algebra, probability & statistics, differential equations, abstract algebra and more), physics, chemistry, biology, computer science.

Levels

Undergraduate core, with graduate-level extensions.

Difficulty

Four tiers (Easy · Medium · Hard · Extreme), ready for curriculum learning and difficulty-balanced RL.

Diversity

Similarity-based deduplication removes repetitive templates and preserves varied reasoning patterns.

04Schema

Record format.

FieldTypeDescription
idstringStable unique identifier.
subjectstringDiscipline, e.g. Mathematics, Physics, Chemistry.
levelstringUndergraduate · Graduate · Competition.
qtypestringProof · Calculation · Short answer · Single / multiple choice · Fill-in-the-blank · True / false.
questionstringProblem statement in Markdown with LaTeX.
optionsarray | nullAnswer options for choice questions.
answerstringFinal answer, verifiable where the question type allows.
solutionstringFull step-by-step reasoning.
sub_disciplinestringe.g. Calculus, Linear Algebra, Abstract Algebra.
difficultyenumEasy · Medium · Hard · Extreme.
05Use

Built for.

  • Reasoning SFTComplete, structured solutions for multi-step scientific reasoning.
  • RL with verifiable rewardsCalculation and choice questions with checkable final answers.
  • General logical reasoningEven general-purpose models build reasoning through problem-based RL.
  • Curriculum learningDifficulty tiers map directly to progressive training schedules.
Contact

Request a pilot slice.

Tell us the undergraduate stem slice you need. We will send a pilot sample cut to your specification for your own evaluation.