Reproduce “Evaluating Fuzz Testing for Reinforcement Learning Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhibin Kang, Hanmo You, Dong Wang, Haiming Zheng, Junjie Chen
The authors compare five reinforcement-learning fuzzers and random testing across MountainCar, BipedalWalker, and CARLA under a standardized evaluation. They measure crash discovery, diversity, efficiency, and downstream use. In a 12-hour BipedalWalker test, MDPFuzz found 11,945 mean valid unique crashes versus 8,026 for random testing, but methods differed in the failures they uncovered; downstream evidence is limited to BipedalWalker.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.