04:00
2026-07-24
arxiv.org
artificial-intelligence
SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification
SonicSampler, a unified suite of tile-aware Triton kernels from arXiv, vertically fuses the complete LLM sampling pipeline into a fixed, workload-aware execution model, achieving up to 16x speedup ove…