Robustness on GAIA
How well does the agent maintain accuracy when inputs are perturbed (faults, structural changes, prompt rewording)? See full definition →
Sub-metric Comparison
Baseline vs Perturbed Accuracy
Agent Leaderboard — Robustness
| # | Agent | Acc | Reliability | Robustness Agg | Fault | Struct | Prompt |
|---|---|---|---|---|---|---|---|
| 1 |  O1 |
34.4% | 0.79 | 1.00 | 1.00 | 1.00 | 1.00 |
| 2 |  Claude Opus 4.7 |
73.3% | 0.84 | 0.99 | 1.00 | 1.00 | 0.96 |
| 3 | GPT-4o Mini |
26.3% | 0.76 | 0.97 | 1.00 | 1.00 | 0.92 |
| 4 |  Gemini 3.5 Flash |
79.2% | 0.84 | 0.96 | 1.00 | 1.00 | 0.88 |
| 5 | GPT-5.5 |
62.8% | 0.79 | 0.95 | 1.00 | 0.96 | 0.91 |
| 6 | Claude Sonnet 4 |
54.7% | 0.82 | 0.94 | 1.00 | 1.00 | 0.83 |
| 7 | Claude 3.5 Haiku |
25.7% | 0.82 | 0.94 | 1.00 | 1.00 | 0.82 |
| 8 | Gemini 3.1 Pro |
76.2% | 0.82 | 0.94 | 0.99 | 0.96 | 0.86 |
| 9 | GPT-5.2 (medium) |
31.8% | 0.72 | 0.91 | 1.00 | 1.00 | 0.74 |
| 10 | Gemini 2.5 Pro |
52.5% | 0.78 | 0.91 | 0.95 | 0.97 | 0.82 |
| 11 | Claude Opus 4.5 |
68.5% | 0.85 | 0.91 | 0.99 | 0.98 | 0.77 |
| 12 | Gemini 2.5 Flash |
46.7% | 0.74 | 0.87 | 0.97 | 0.96 | 0.69 |
| 13 | GPT-4 Turbo |
30.8% | 0.76 | 0.87 | 1.00 | 0.91 | 0.71 |
| 14 | Claude 3 Haiku |
12.9% | 0.69 | 0.86 | 1.00 | 0.94 | 0.66 |
| 15 | GPT-5.2 |
33.2% | 0.72 | 0.78 | 0.62 | 1.00 | 0.73 |