ToolHop
Fits research or lookup agents that must chain calls where each step depends on the last. Borrow the query-driven method of writing the tools after the question so every case has a verifiable answer.
What it asks the agent to do
- answer a question that needs three or more dependent tool calls
- pick the next tool from the result of the previous call
- stop and give a verifiable final answer
What to borrow from it
case templatebaseline
Caveats
One user query per case with no conversation; questions derive from MuSiQue, and the code lives in a Hugging Face dataset repo rather than a maintained GitHub project.
Details
| Workloads | knowledge qa, tool calling, research |
|---|---|
| Interaction | single turn |
| Environment | api sandbox |
| End state checked | No |
| Tools | 3,912 locally executable code tools generated per query |
| Grading | exact match |
| Size | 995 |
| Data | mixed |
| Personal data | none |
| License | Apache-2.0 |
| Harness | own repo |
| Status | active |