WEFT: Scaling Tool-Use Post-Training for General-Purpose Agents
Open paper
Benchmark
Model
Rank
Results
agents-on-bench
WEFT-35B-A3B
–
Accuracy: 74.34
agents-on-bfclv4
WEFT-14B
–
Accuracy: 67.63
agents-on-claw-eval
WEFT-35B-A3B
–
Accuracy: 64.0
agents-on-toolathlon-verified-108-tasks
WEFT-35B-A3B
–
Pass@1: 0.4599
Rank counts only results with a code link.