benchmark-radar
by ktwu01
Catalog of 9,000+ benchmarks, datasets, and eval records with linked evidence
How It Works
Aggregates and tracks 9,000+ AI benchmark, evaluation, dataset, and data-quality records from 37 public sources with daily updates. Normalizes links, evidence, and metadata so you can search and compare benchmark results and dataset provenance across sources. Distinctive for its breadth of indexed records and continuous refresh cadence that surfaces dataset lineage and evaluation artifacts in one catalog, leveraging semantic capability matching.
The Value Proposition
When to Use
Researchers and engineering teams who need a broad, continuously updated index of benchmarks and dataset provenance to inform agent evaluation and trust decisions, aligned with an MCP-backed workflow for context-rich evaluation, i.e. Model Context Protocol (MCP).
How It's Used
- Compare model performance across multiple benchmark suites to validate agent claims
- Trace dataset provenance and data-quality signals before using a benchmark for agent evaluation
- Feed aggregated benchmark and dataset evidence into agent reputation or RepKit-style systems
- Monitor new benchmark releases and dataset updates that could affect agent validation