benchmark + agent
BrowserGym
A framework and benchmark family for evaluating web agents in browser-based environments.
What it is
BrowserGym supplies unified browser observation and action spaces plus integrations with multiple web task benchmarks.
Related records
SWE-agentA research system that enables language models to resolve software engineering tasks in repositories.WebArenaA realistic web environment for evaluating autonomous agents on reproducible tasks.Inspect AIAn open-source framework for building and running language-model evaluations.OpenAI EvalsAn open-source framework and registry for evaluating language models and model systems.