Mahal Systems

Overview › Benchmark catalog

ShoppingBench

Fits shopping assistants with budget or promotion logic. Borrow the intent-grounded instruction generation from sampled products.

What it asks the agent to do

  • find or buy a product in a simulated store with 2.5 million real items
  • apply a voucher, stay within a budget, find a seller with several wanted items

What to borrow from it

user simulatorcase template

Caveats

Task count, code and license not found; instructions are simulated; GPT-4.1 under 50 percent at release.

Details

Workloadsshopping, retail
Interactionsingle turn
Environmentapi sandbox
End state checkedYes
Toolssimulated shopping platform
Gradingprogrammatic check
SizeNot published
Datamixed
Personal datanone
Licenseunknown
Harnessnone
Statusactive

Links