math benchmark
TheoremQA
A theorem-driven question answering dataset containing 800 high-quality questions covering 350+ theorems from Math, Physics, EE&CS, and Finance. Designed to evaluate AI models' capabilities to apply theorems to solve challenging university-level science problems.
Updated Aug 7, 2026
Published models6
Registry coverage6
MetricScore
EvidenceB
TheoremQA leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 44.4% | 100.0% | 6 | C | |
| 2 | AC | 44.1% | 80.0% | 6 | C | |
| 3 | AC | 43.1% | 60.0% | 6 | C | |
| 4 | AC | 43.0% | 40.0% | 6 | C | |
| 5 | AC | 34.0% | 20.0% | 6 | C | |
| 6 | AC | 25.3% | 0.0% | 6 | C |
Score distribution
Top published rows on the benchmark's original scale.
TheoremQA highlights
The top published results on this benchmark's own scale.
What is TheoremQA?
Definition and scoring fields from the benchmark registry.
A theorem-driven question answering dataset containing 800 high-quality questions covering 350+ theorems from Math, Physics, EE&CS, and Finance. Designed to evaluate AI models' capabilities to apply theorems to solve challenging university-level science problems.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- TheoremQA
- Modality
- text
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- theoremqa|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about TheoremQA.
Which model scores highest on TheoremQA?
Qwen2 72B Instruct is currently ranked first with 44.4%.
What does TheoremQA measure?
A theorem-driven question answering dataset containing 800 high-quality questions covering 350+ theorems from Math, Physics, EE&CS, and Finance. Designed to evaluate AI models' capabilities to apply theorems to solve challenging university-level science problems.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
6 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.