Eval directory · Data Analysis
Evaluation packs covering adversarial robustness, safety gates, workflow quality, and operator-level checks for Databricks AI products.
Databricks is the Data + AI Company, providing a unified lakehouse platform for data engineering, analytics, and machine learning at scale. Thousands of organizations use Databricks to build, train, and deploy AI and ML workloads on their own data.
Use the eval library for Databricks
All 3 test cases — inputs, expected behavior, and pass/fail checks — runnable in Corsac with your own data.
Related in Data Analysis
All evals →Coverage map
1 area · 3 graded scenarios
Every eval set is graded on
Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls
01
Evaluates Databricks' Assistant NL2SQL — query intent translation, schema-constrained generation, and result reliability — across 3 test cases graded case by case by an LLM judge.
Mapped capabilities
3 scenarios
Example criterion: Databricks reliably translates analyst questions into accurate SQL that returns trustworthy business results.
Each eval pack tests Databricks's public product surface — including Assistant Nl2sql Smoke V1 — against graded scenarios covering adversarial robustness, workflow quality, safety gates, and operator quality. Every pack is runnable in Corsac with your own data.
Every eval set is graded the same way: pass/fail checks plus an LLM judge scoring 1–5 against each case's expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.
The Databricks eval library includes 3 graded test cases across 1 eval pack. Each case defines an input, expected behavior, and pass/fail criteria.
Request the library with your work email above and we'll set it up in a Corsac workspace, where you can run every test case against Databricks or your own agent with your own data.
