NuminaMath-TIR Math Problems
Source, license and coverage
Supplier documentation. These claims are separate from the automated sample score. A listing edit date is not a data freshness date.
- License
- apache-2.0
- Source / creator
- AI-MO/NuminaMath-TIR
- Collection method
- The authors selected ~70k problems from NuminaMath-CoT, restricting to problems with numerical (mostly integer) answers to enable automatic verification. They then ran a GPT-4-based pipeline that produces TORA-style reasoning trajectories: the model reasons in natural language, emits Python code, the code is executed, and the outputs are fed back into the trajectory. Trajectories where the final numerical answer matched the ground truth were retained. The resulting traces simulate how a tool-using math agent should think and call a code interpreter.
- Coverage start
- Not documented
- Coverage end
- Not documented
- Data last updated
- Not documented
- Update schedule
- Not documented
- English-only; no multilingual coverage. - Restricted to problems with numerical (mostly integer) final answers — excludes proofs, symbolic answers, and open-ended problems. - Reasoning traces are GPT-4 synthetic, not human-authored — may inherit GPT-4 biases, stylistic patterns, and occasional subtle errors even when the final answer is correct. - Source problems originate from competition and textbook corpora; overlap with public math benchmarks (MATH, GSM8K, AIME) is possible and not explicitly deduplicated. - Source does not document additional gaps; buyers should validate empirically before training.
Sample structure score: 100 / 100
This automated check describes the inspected sample, not factual accuracy, legal rights, representativeness, or the quality of the entire dataset. It is not a customer rating.
Assessed 10 sample records (JSON) on 2026-10-09. All records in the provided sample were checked.
| Check | Points | Evidence |
|---|---|---|
| Populated cells | 50 / 50 | 30 of 30 top-level cells contain a value. Null, absent and blank values count as missing; zero and false count as populated. |
| Consistent value types | 30 / 30 | 30 of 30 populated cells match their column's most common observed type. Types are inferred, not checked against real-world truth. |
| Consistent record shape | 20 / 20 | 10 of 10 records have the expected fields. CSV/TSV use the header width; JSON uses the union of observed keys. |
Field-level findings and improvements
Check missing cells and mixed types below. Document intentional missing values or mixed types in your field descriptions. Do not fill legitimate unknowns with invented values just to increase this score.
| Field | Missing cells | Most common type | Other populated types |
|---|---|---|---|
| problem | 0 / 10 | string | 0 / 10 |
| solution | 0 / 10 | string | 0 / 10 |
| messages | 0 / 10 | object | 0 / 10 |
About this data
Math problems with tool-integrated reasoning traces generated via GPT-4, derived from NuminaMath-CoT. Designed for training math-reasoning agents with code execution capability.
Retrieve with your agent or Python
Create an account and configure DATABAZAAR_API_KEY. This example retrieves free or already purchased data; it never makes a purchase. For a multi-file dataset, choose a file index from the manifest.
Download the Python examplepython3 retrieve-dataset.py a5ad6289-6fc8-4079-94de-b2dad1f0388f --output dataset.bin
Full supplier documentation
Schema
| Name | Type | Description |
|---|---|---|
| problem | VARCHAR | English-language mathematics problem statement |
| solution | VARCHAR | Tool-integrated reasoning trace with interleaved natural language, executable Python code blocks, and their outputs |
| messages | STRUCT("content" VARCHAR, "role" VARCHAR)[] | Chat-formatted message list with role (system/user/assistant) and content fields for supervised fine-tuning |
Sample Data
Preview a sample of the data before downloading.
Public sample only. Sign in to retrieve the full dataset, including free datasets.
For AI Agents
# 1. Add to your agent's MCP config (claude_desktop_config.json or similar):
{
"mcpServers": {
"databazaar": { "command": "npx", "args": ["databazaar-mcp"] }
}
}
# 2. Your agent can then call:
search_datasets({ query: "NuminaMath-TIR Math Problems" })
// Found: a5ad6289-6fc8-4079-94de-b2dad1f0388f
get_download_url({ dataset_id: "a5ad6289-6fc8-4079-94de-b2dad1f0388f" }) // free — sign in with MCP OAuth first# Free dataset — sign in or use your account API key: curl https://api.databazaar.io/datasets/a5ad6289-6fc8-4079-94de-b2dad1f0388f/download-url -H "Authorization: Bearer $DATABAZAAR_API_KEY"