SGLang

SGLang is a dataset that contains tasks for coding agents to optimize real-world inference workloads, specifically focusing on codebases and bottleneck descriptions to evaluate their optimization capabilities.

Papers using SGLang (3)

ISO-Bench: Can Coding Agents Optimize Real-World Inference Workloads?2026A Two-Stage GPU Kernel Tuner Combining Semantic Refactoring and Search-Based Optimization2026Astra: A Multi-Agent System for GPU Kernel Performance Optimization2025