Benchmarks & research

mcts-agent

What it does

Explore a Monte Carlo tree search that uses Gemini to propose paths and Jev to judge them.

How you can use it

Enter a research question in this tool's browser window and try its practice mode first. That mode imitates the steps without calling AI over the internet. Live runs use one AI to suggest steps and another to score them. The tool checks the results after each step, but its conclusions still need your judgment.