Claude Sonnet 4.5
on τ-bench (airline, original)
54.4%
Accuracy
0.79
Overall Reliability
#2
of 13 agents
0.73
Consistency
0.66
Predictability
0.98
Robustness
0.97
Safety
| Dimension | Agg | Sub-metrics |
|---|---|---|
| Consistency | 0.735 |
Outc
0.540
Traj-D
0.865
Traj-S
0.769
Res
0.847
|
| Predictability | 0.663 |
Cal
0.678
AUROC
0.615
Brier
0.663
|
| Robustness | 0.980 |
Fault
0.972
Struct
1.000
Prompt
0.968
|
| Safety | 0.970 |
Harm
0.500
Comp
0.940
Safety
0.970
|
Per-Task Outcome Consistency
Each cell represents a task. Color shows outcome consistency across runs. Hover to see task ID.
Consistency Distribution
KDE of per-task outcome consistency. Peaks at 0 or 1 indicate polarized behavior.
Per-Task Cost Distribution
KDE of mean cost per task (averaged across runs).
Per-Task Time Distribution
KDE of mean execution time per task (averaged across runs).
Calibration Curve
Risk-Coverage (AURC)
Confidence Distribution
Distribution of expressed confidence values across tasks.