cs.LGNov 18, 2026

RSIBench-Infra: Can Agents Optimize Their Own Training and Inference Infrastructure at Production Scale?

Authors: Hanlin Wu, Yifei Wang, Zhicheng Cai, Huan-ang Gao, Wei-Ying Ma, Ya-Qin Zhang, Jingjing Liu, Hao Zhou

Organizations: ByteDance Seed · Institute for AI Industry Research (AIR), Tsinghua University · SIA-Lab of Tsinghua AIR and ByteDance Seed

Abstract

Improving the efficiency of training and inference infrastructure is a practical component of recursive self-improvement. Existing benchmarks for agent-driven optimization largely focus on individual software layers such as kernel, leaving unclear whether agents can deliver end-to-end performance gains under production workloads. We introduce RSIBench-Infra, a benchmark of 20 training and inference tasks spanning diverse model families and modalities, with deployments of up to 32 accelerators on NVIDIA GPUs and Ascend NPUs. Agents start from runnable baselines derived from official recipes and can modify the software stack across engine, kernel, and instruction levels. An OCI image-layer submission protocol captures these changes for reproducible deployment, while independent evaluation measures end-to-end performance under task-specific quality constraints. Across all task results, 66.2% exceed 1.10×1.10\times validated speedup. These results demonstrate that current agents can improve production-scale training and inference performance, but identifying effective optimization strategies and consistently translating local improvements into end-to-end gains remain challenging.

Explore similar work

CardsList
  1. Infrathon: Benchmarking Language Agents for Real-World, Long-Horizon AI Infra Optimization

    Nov 9, 2026Zhengyu Hu, Zheyuan Xiao, Yike Wang +18Long-Horizon Agent EvaluationAI Agent Benchmarks

  2. RSI-Opt: Recursive Self-Improvement Optimizer Agents For Model Training

    Oct 23, 2026Binghang Lu, Yige Sun, Guang LinLarge Language Model-Guided OptimizationLearning to Optimize