Mahal Systems

Overview › Benchmark catalog

TravelPlanner

Fits planning assistants that must juggle many constraints. Borrow the split between commonsense and hard constraint pass rates plus a final all-constraints pass rate.

What it asks the agent to do

  • produce a full multi-day itinerary from a request
  • look up flights, hotels, restaurants and attractions with tools
  • satisfy hard constraints (budget, cuisine, room rules) and commonsense ones

What to borrow from it

grading designcase templatebaseline

Caveats

Static offline data; original pass rates were near zero so numbers are hard to compare across versions; no interaction with a user.

Details

Workloadstravel, planning, tool calling
Interactionsingle turn
Environmentapi sandbox
End state checkedNo
Toolssearch tools over a static database of about four million records
Gradingprogrammatic check
Size1,225
Datamixed
Personal datanone
LicenseMIT
Harnessown repo
Statusactive

Links