Back to code-only MCP servers

ground-truth-evals

active

An LLM eval harness graded by computed ground truth, not an LLM judge. Worked example: poker, served to models as MCP tools — gpt-4o-mini goes from 26/75 to 75/75 with them.

Code-only MCP server: no hosted endpoint, so you install and run it yourself. It is not probed, graded or counted in our reachability statistics.

Get it

Repository
github.com/yangstar89/ground-truth-evals
License
MIT
Last update
2026-09-23
GitHub stars
★ 1

Listed in

Run this server as a hosted endpoint? Submit the endpoint and it moves to the remote directory, where it is probed and graded.