Mahal Systems WebChoreArena
Fits back-office browser agents that must aggregate data across many pages. Borrow the memory / calculation / long-term-memory task categories.
What it asks the agent to do
- read and total many records across pages (orders, posts, issues)
- compute a figure from information gathered on the site
- track details across many pages and sites to answer
What to borrow from it
case templatebaseline
Caveats
Needs the WebArena site stack; top reported score 44.9 percent on a preview model since retired; cross-site tasks are few.
Details
| Workloads | web browsing, retail, devops, data analysis |
|---|---|
| Interaction | single turn |
| Environment | browser |
| End state checked | Yes |
| Tools | browser actions on WebArena sites |
| Grading | exact match, programmatic check, state diff |
| Size | 532 |
| Data | mixed |
| Personal data | possible |
| License | Apache-2.0 |
| Harness | own repo |
| Status | active |