OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web Agents
Open paper
Benchmark
Model
Rank
Results
computer-use-agents-on-online-mind2web
OpenWebRL-8B
#4
Success Rate: 69.7
computer-use-agents-on-webvoyager
OpenWebRL-8B
#7
Success Rate: 74.6
Rank counts only results with a code link.