AG News Topic Classification Benchmark
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
- unknown
- Source / creator
- fancyzhx/ag_news
- Collection method
- ComeToMyHead aggregated more than one million news articles from 2000+ sources over more than a year of crawling starting in 2004. Zhang et al. selected the 4 largest topic classes from this corpus and sampled 30,000 training and 1,900 test examples per class to produce the balanced benchmark distribution that is mirrored here. Text is the raw title + description concatenation with minimal normalization.
- Coverage start
- Not documented
- Coverage end
- Not documented
- Data last updated
- Not documented
- Update schedule
- Not documented
Articles date from roughly 2004-2005, so topical vocabulary, named entities, and world events are stale relative to modern news. English-only. The 4-class taxonomy is coarse and collapses many subtopics. License is declared "unknown" on the HF card — buyers redistributing commercially should confirm acceptable use with the original ComeToMyHead terms. Likely heavy overlap with pretraining corpora of most modern LLMs, so zero-shot numbers on AG News overstate generalization.
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 | 20 of 20 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 | 20 of 20 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 |
|---|---|---|---|
| text | 0 / 10 | string | 0 / 10 |
| label | 0 / 10 | number | 0 / 10 |
About this data
News articles classified into four topics: World, Sports, Business, and Science/Technology. Standard benchmark for evaluating text classification models and NLP systems.
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 f14329c0-7c5b-4f9f-9c43-3d2a5ef71b1a --output dataset.bin
Full supplier documentation
Schema
| Name | Type | Description |
|---|---|---|
| text | VARCHAR | Concatenated news article title and description text. |
| label | BIGINT | Topic class: 0=World, 1=Sports, 2=Business, 3=Sci/Tech. |
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: "AG News Topic Classification B" })
// Found: f14329c0-7c5b-4f9f-9c43-3d2a5ef71b1a
get_download_url({ dataset_id: "f14329c0-7c5b-4f9f-9c43-3d2a5ef71b1a" }) // free — sign in with MCP OAuth first# Free dataset — sign in or use your account API key: curl https://api.databazaar.io/datasets/f14329c0-7c5b-4f9f-9c43-3d2a5ef71b1a/download-url -H "Authorization: Bearer $DATABAZAAR_API_KEY"