What AI says about Latency-Aware MAS (LAMaS)

LAMaS is a Python-based project that reproduces experiments from the arXiv paper 2601.10560. It provides setup guidance, API key configuration, and a dataset workflow for GSM8K, HumanEval, and MATH, plus a configurable experiment runner that can use models like gpt-4o-mini with tunable parameters such as token weights, token penalties, and learning rate. The repository includes scripts for organizing results and acknowledges MaAS for providing foundational code and prompts.