MetaKernelBench: Measuring GPU Kernel Knowledge Transfer Beyond Code

• 类型：arxiv
• 标识：2610.05014
• 链接：https://arxiv.org/abs/2610.05014
• 主分类：agent
• 形态：benchmark
• TLDR：Recent GPU kernel optimization agents retain what they learn in knowledge bases or as distilled skills. Kernel benchmarks score each attempt's implementation for correctness and speed but leave the reuse value of retained experience unmeasured. We introduce MetaKernelBench, which measures whether experience distilled from an attempt in one kernel domain-specific language (DSL) improves a fresh attempt at the same problem in another. Its 74 problems are fused subgraphs in six families, each posed as a pair of CuTe DSL and TIRx variants that differ only in the DSL. The agent first attempts each
• 待LLM分类：否
• 标题中文：MetaKernelBench：超越代码的 GPU Kernel 知识迁移度量
• TLDR中文：近期 GPU Kernel 优化智能体会将所学内容保留在知识库中或提炼为技能。Kernel 基准对每次尝试的实现在正确性与速度上打分，但未度量所保留经验的复用价值。我们提出 MetaKernelBench，用于衡量从一种 Kernel 领域特定语言（DSL）的尝试中提炼的经验是否能在另一种语言下提升对同一问题的新一轮尝试。其 74 个问题为六类融合子图，每类以仅 DSL 不同的 CuTe DSL 与 TIRx 变体对形式呈现。智能体首先尝试……
• 来源文件：
• /inbox/tom/_candidates/2026-10-12-agent-rag-longcontext-candidates.json