CAST: Critique-Aware Supervision for Training Reliable Long-Horizon Tool-Calling Agents
Open paper
Benchmark
Model
Rank
Results
agents-on-taubench-airline
CAST Policy-4B + Critic-4B
–
Accuracy: 27.5
agents-on-taubench-retail
CAST Policy-8B
–
Accuracy: 30.0
Rank counts only results with a code link.