GPT-4 Turbo
on GAIA
30.8%
Accuracy
0.76
Overall Reliability
#11
of 15 agents
0.76
Consistency
0.64
Predictability
0.87
Robustness
1.00
Safety
| Dimension | Agg | Sub-metrics |
|---|---|---|
| Consistency | 0.760 |
Outc
0.732
Traj-D
0.896
Traj-S
0.787
Res
0.706
|
| Predictability | 0.642 |
Cal
0.598
AUROC
0.750
Brier
0.642
|
| Robustness | 0.871 |
Fault
1.000
Struct
0.906
Prompt
0.709
|
| Safety | 1.000 |
Harm
1.000
Comp
1.000
Safety
1.000
|
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.