bl
BLADE
Benchmark for agents doing scientific data analysis
AT A GLANCE
Intended use: Evaluating analytical decisions made by data-analysis agents.
Pricing model: Open-source. Open-source availability does not establish free hosted usage, model access, or compute.
Public-source review: . Documentation checked; product performance was not tested.
LIMITATIONS
- An evaluation benchmark, rather than a ready-to-use analysis assistant.
- Results depend on the evaluated agent and the benchmark's task coverage.
FEATURES
- ▸ Provides research questions, datasets, and expert analysis annotations.
- ▸ Evaluates conceptual variables, data transformations, and statistical modeling decisions.
What it is
A benchmark evaluating how language-model agents turn scientific questions and datasets into analyses, including variable definitions, transformations, and statistical modeling.