Research-level multimodal math records spanning 18 domains with 143 referenced figures in the sample, built for the hardest MMMU-Pro, OlympiadBench multimodal math, and MathVision-style problems.
Available Now · Sample Available
Multimodal Math QSA
The hardest multimodal math in the OTS line — MMMU-Pro, OlympiadBench MM, and MathVision territory.
Volume
0
Volume
Scalable Volume
Coverage
Grad, PhD, Expert, Math, Mathematics
17.4%Frontier Model · Pass@1
40.4%Claude Opus 5 · Pass@1
46.8%Claude Opus 5 · Pass@3
Targets
MMMU (Math), MathVista, OlympiadBench (MM math), MMMU-Pro, MathVision
Modality & Delivery
ImageText→Text
Formats
Excel (.xlsx)Images
What You Receive
Coverage
OTS’s most difficult multimodal math product. The source contains 1,000 research-level records built for the hardest MMMU-Pro, OlympiadBench multimodal math, and MathVision-style problems. The sample contains 50 records across 18 domains with 143 referenced figures.
Package Structure
Differentiators
- Research-level multimodal mathematics with dense figure use across eighteen domains in the sample.
Record Schema
| A | B | C | D | |
|---|---|---|---|---|
| 1 | answer | image_ref | question | solution |
| 2 | The concise final answer associated with the question. | Path to the associated image file referenced by the record. | The question or task presented in the dataset record. | The worked solution or supporting reasoning for the question. |
Results
| Model | Metric | Result |
|---|---|---|
| Frontier Model | Pass@1 | |
| Claude Opus 5 | Pass@1 | |
| Claude Opus 5 | Pass@3 |
Methodology & Quality
Methodology
SME involvement
Domain-expert curated with up to L2 review
Stressors
Multi-figure mathematical interpretation at research difficulty.
Design criteria
- Target the hardest MMMU-Pro, OlympiadBench multimodal math, and MathVision-style tasks.
Build assumptions
Quality approach
Known limitations
Customization & Scalability
Custom sets available. Additional volume in progress
Get Started
Request a Sample
Share your target domains and the OTS team will follow up with sample data and next steps.