Tool Calling on WHEN2TOOL Easy
-0.3ΔAccNecessary (N)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Necessary (N)model averaging=six models2026.05 | -0.3 | -0.09 | -3.5 | |
| PROBE&PREFILLmodel averaging=six models, threshold (τ)=0.52026.05 | -1.1 | -0.66 | -1.6 | |
| Sparse (S)model averaging=six models2026.05 | -6.3 | -0.55 | -11.3 | |
| Necessary + Reason-then-Actmodel averaging=six models2026.05 | -8.1 | -0.95 | -8.5 | |
| Sparse + Reason-then-Actmodel averaging=six models2026.05 | -9.9 | -1.13 | -8.8 | |
| No Tool + Reason-then-Actmodel averaging=six models2026.05 | -12.4 | -1.18 | -10.5 | |
| No Tool (X)model averaging=six models2026.05 | -18.1 | -0.72 | -25.2 |