Research-grade algorithmic reasoning that leading frontier models still fail on 37% of the time. Each record pairs a question with a worked solution and a final answer.
Available Now · Sample Available
Python Coding QSA
PhD-level Python: ML pipelines, numerical methods, and cryptographic protocols.
Volume
0
Volume
Scalable Volume
Coverage
Grad, PhD, Python, Coding
Model Benchmark
0.0
%
Codex GPT-5.4 · Pass at k
Targets
LiveCodeBench, HumanEval, MBPP, BigCodeBench
Modality & Delivery
Text→Text
Formats
Excel (.xlsx)
What You Receive
Coverage
ML pipelines, numerical methods, cryptographic protocols, and bioinformatics algorithms — suited to fine-tuning on algorithmic reasoning rather than repo-patch tasks.
Package Structure
Differentiators
- Structured question, worked solution, and final answer record format.
Record Schema
| Field | Description |
|---|---|
answer | The concise final answer associated with the question. |
question | The question or task presented in the dataset record. |
solution | The worked solution or supporting reasoning for the question. |
Results
| Model | Metric | Result |
|---|---|---|
| Codex GPT-5.4 | Pass@3 |
Methodology & Quality
Methodology
Questions are expert-authored, then split across an SME pool for complexity validation before benchmarking.
SME involvement
Domain-expert curated with up to L2 review
Stressors
Design criteria
- Keep the question, supporting solution, and final answer as distinct record components.
Build assumptions
Quality approach
Known limitations
Customization & Scalability
Custom sets available. Additional volume in progress