Mahal Systems ALFWorld
Fits teams who need a cheap, deterministic multi-step environment for smoke tests of a planning loop; it is also the base of the household tasks in AgentBench, AgentBoard and AgentGym.
What it asks the agent to do
- find an object and put it somewhere in a simulated house
- heat, cool or clean an object before placing it
- examine an object under a light
What to borrow from it
sandbox designbaseline
Caveats
From 2020 and reported near ceiling by many recent agent papers; a text game, not a work task; task counts were not visible on the README.
Details
| Workloads | embodied, planning |
|---|---|
| Interaction | multi turn scripted |
| Environment | api sandbox |
| End state checked | Yes |
| Tools | text world action interface |
| Grading | programmatic check |
| Size | Not published |
| Data | synthetic |
| Personal data | none |
| License | MIT |
| Harness | own repo |
| Status | saturated |