multiverse-parallel-generation

Enable native parallel token generation in language models by implementing adaptive task decomposition and merge strategies, achieving 2x speedup with 1.87% performance gains.

Independent GitHub snapshot · Automated classification · Original documentation governs