Turing Open Reasoning: STEM QA for Multi-Step Symbolic Manipulation

TuringEnterprises published this collection of fewer than 1,000 STEM reasoning problems in December 2025 to benchmark deep reasoning in large language models. The data covers Physics, Mathematics, Biology, and Chemistry through tasks requiring symbolic manipulation, numerical accuracy, and simulation-based verification.

Use Cases

• Evaluating LLM performance on multi-step symbolic manipulation in Physics
• Benchmarking numerical accuracy in Chemistry and Biology problems
• Testing simulation-based verification capabilities of reasoning models

Strengths

• Covers four distinct STEM domains: Physics, Math, Biology, and Chemistry
• Includes problems specifically designed for simulation-based verification
• Released under the MIT license for open research

Limitations

• Small sample size of fewer than 1,000 records
• Specific column schema and field names are not documented in the primary metadata

Provenance

Source

TuringEnterprises

Freshness

Last updated December 2025.