Evolving Agent Teams (EAT) is a multi‑agent framework designed to overcome local optima that plague existing coding agents when optimizing GPU kernels. It employs a central supervisor to diagnose and restructure agent teams and a persistent file system that records learning documents, evolution history, and team artifacts. Across a range of benchmarks—including FlashInfer, FlashKDA, and RoboDojo—EAT consistently outperforms strong coding‑agent baselines, achieving up to 1.2× speedup on FlashInfer kernels and superior results on several other tasks.
arXiv:2608. 12629v1 Announce Type: new Abstract: GPU kernel agents and GPU programming languages have advanced separately, leaving expert kernels difficult to reproduce.
KernelArc is a multi-agent framework designed to autonomously optimize GPU kernels across diverse workloads. It employs strategy-specialized agents that run concurrently, coordinating via conclusions-only shared memory, a deterministic benchmark guard, and read-only cross-agent state with plateau-triggered drafting. Evaluated on NVIDIA H100 and B200 GPUs with SOL-ExecBench workloads, KernelArc produced top-ranked implementations for tasks such as BF16 GEMM, cuBLASLt configuration tables, and various attention mechanisms, achieving first place on several leaderboard categories.
InfraDev is a new benchmark comprising 36 tasks derived from real-world pull requests to Megatron‑LM, vLLM, and verl, designed to test coding agents on infrastructure development for training, inference, and reinforcement learning across many GPUs. Each task requires agents to reimplement multi‑file changes and run correctly, with grading that includes hidden probes and a prompt exposing only the interfaces used. The benchmark reveals that current frontier models resolve only about 39% of tasks, with probes rejecting a third of submissions that pass official tests, and that structural and state errors dominate failures.
arXiv:2607. 24762v1 Announce Type: new Abstract: Machine learning models are increasingly embedded in everyday software, and most of their runtime is spent in a small set of compute kernels such as matrix multiplication, convolution, and normalization.
arXiv:2607. 04395v1 Announce Type: new Abstract: Recent agentic approaches to LLM-based kernel generation have achieved impressive results on CUDA.
arXiv:2604. 01489v2 Announce Type: replace Abstract: High-performance GPU kernels are critical to modern machine learning systems, yet developing them remains a manual, expert-driven process.
The article describes a case study where a large language model (LLM) agent loop was used to optimize three Gemma‑4‑12B decode kernels for the FuriosaAI RNGD NPU, a sparsely documented Rust DSL accelerator. Over 12 days on a shared device, the loop built its own target knowledge, performed microbenchmarks, recorded compiler legality rules, and iteratively refined kernel designs through role‑based search and acceptance tests. The resulting kernels achieved a 9.293× geometric mean speedup over the official baseline, each within 1.25× of the theoretical cycle minimum for weight movement.
FastKernels is a benchmark comprising 384 GPU kernel tasks from 47 architectures across 8 categories, designed to evaluate kernel generation agents in realistic production settings. It covers 94.6% of HuggingFace Transformers architectures, scoring kernels on correctness, coverage, and speedup within their native execution paths. The benchmark reveals that kernel‑level speedups often collapse to modest end‑to‑end gains and that isolated performance rankings can misrepresent real‑world effectiveness.
arXiv:2609.17391v1 Announce Type: new Abstract: Model serving is one of the largest cost drivers in production recommender systems. Maximizing its throughput requires navigating a deeply layered hier...
arXiv:2604.18616v2 Announce Type: replace-cross Abstract: LLM coding agents can generate correct GPU kernels, but their performance still trails expert libraries. Reaching peak throughput requires co...
MaxKernel is a multi‑agent system designed to generate high‑performance custom kernels for TPUs. It offers three paradigms: a Human‑in‑the‑Loop agent for collaborative design, an Autonomous agent that runs a fully automated optimization loop, and a Graph‑Based Autonomous Search for global exploration. All paradigms share specialized sub‑agents for planning, implementation, debugging, testing, and profiling, and the system consistently matches expert hand‑tuned baselines on the JaxBench suite and real‑world workloads.
arXiv:2607. 03451v1 Announce Type: cross Abstract: While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines.