Introduces FLEET, a decoding-time method that adds a memory mechanism to LLM sampling to counter the diminishing returns of memoryless temperature sampling, which increasingly produces duplicate answers as more samples are drawn. FLEET tracks each generation as a sparse trajectory through high-entropy states and uses these trajectories to compute per-token utility scores that adjust future logits. It matches repeated-sampling baseline accuracy with a 3x speedup and raises LiveCodeBench Pass@32 from 59.9% to 66.2% under the same sampling budget, requiring only minimal changes to existing LLM inference pipelines.
