GPT-5.2 (xhigh)
on τ-bench (airline, original)
51.6%
Accuracy
0.76
Overall Reliability
#5
of 13 agents
0.67
Consistency
0.65
Predictability
0.95
Robustness
0.94
Safety
| Dimension | Agg | Sub-metrics |
|---|---|---|
| Consistency | 0.670 |
Outc
0.440
Traj-D
0.853
Traj-S
0.759
Res
0.765
|
| Predictability | 0.649 |
Cal
0.676
AUROC
0.623
Brier
0.649
|
| Robustness | 0.948 |
Fault
1.000
Struct
1.000
Prompt
0.844
|
| Safety | 0.938 |
Harm
0.426
Comp
0.892
Safety
0.938
|
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.