Implements a distributed Monte Carlo Tree Search system for exploring reasoning paths in parallel. Uses Modal's distributed primitives to spawn workers that explore different thought chains, using UCB1 to balance exploration vs exploitation. Features: - Parallel reasoning path exploration (20 concurrent workers) - UCB1-based tree search algorithm - LLM-powered step generation and evaluation - Example problems: math, logic, code generation
Miscellaneous Examples
This directory contains a variety of examples of ways to use Modal.
Unlike the examples in the rest of this repository, these examples are not continually monitored for correctness, so it is possible that they may become out of date or incorrect over time.
If you find an error in one of these examples, please report it in the issues tab or, even better, submit a pull request to fix it.