Mahal Systems WebArena-Infinity
Fits teams that need many fresh, uncontaminated browser tasks; borrow the pipeline that generates an app, its tasks and verifiers with coding agents and audits them with browser agents.
What it asks the agent to do
- complete generated tasks on auto-built web apps
- difficulty-graded tasks with programmatic verifiers
What to borrow from it
sandbox designcase template
Caveats
Blog post only, no paper; task count not published; environments are synthetic apps so realism is unverified; very new.
Details
| Workloads | web browsing, workflow automation |
|---|---|
| Interaction | single turn |
| Environment | browser |
| End state checked | Yes |
| Tools | browser actions on generated web apps |
| Grading | programmatic check, state diff |
| Size | Not published |
| Data | synthetic |
| Personal data | none |
| License | MIT |
| Harness | own repo |
| Status | active |