FACT is a three-stage agent-driven system that synthesizes and composes CUTLASS kernels from PyTorch modules, achieving up to 2.03x speedup on transformer blocks over PyTorch and competing optimizers.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 2verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
CuLifter recovers types from untyped GPU register files via constraint propagation to lift 99.98% of 24,437 functions across 919 cubins to valid LLVM IR.
citing papers explorer
-
FACT: Compositional Kernel Synthesis with a Three-Stage Agentic Workflow
FACT is a three-stage agent-driven system that synthesizes and composes CUTLASS kernels from PyTorch modules, achieving up to 2.03x speedup on transformer blocks over PyTorch and competing optimizers.
-
CuLifter: Lifting GPU Binaries to Typed IR
CuLifter recovers types from untyped GPU register files via constraint propagation to lift 99.98% of 24,437 functions across 919 cubins to valid LLVM IR.