AgentDojo
AgentDojo: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
28rows
robustness_scoreprimary metric
2026-05-27sampled
Metadata
Metrics
Robustness Score, Utility, Utility under attack, Targeted ASR (lower is better)
| Rank | Subject | Robustness Score | Model Match | Provenance | Sampled |
|---|---|---|---|---|---|
| 1 | claude-3-5-sonnet-20241022 (None, important_instructions) | 85.69% | — | Imported | 2026-05-27 |
| 2 | claude-3-7-sonnet-20250219 (None, important_instructions) | 84.98% | — | Imported | 2026-05-27 |
| 3 | gpt-4o-2024-05-13 (None, direct) | 81.8% | — | Imported | 2026-05-27 |
| 4 | gpt-4o-2024-05-13 (None, injecagent) | 81.4% | — | Imported | 2026-05-27 |
| 5 | gpt-4o-2024-05-13 (None, ignore_previous) | 80.68% | — | Imported | 2026-05-27 |
| 6 | gpt-4o-2024-05-13 (tool_filter, important_instructions) | 74.72% | — | Imported | 2026-05-27 |
| 7 | claude-3-opus-20240229 (None, important_instructions) | 70.59% | — | Imported | 2026-05-27 |
| 8 | gpt-4o-2024-05-13 (repeat_user_prompt, important_instructions) | 69.72% | — | Imported | 2026-05-27 |
| 9 | gemini-1.5-pro-002 (None, important_instructions) | 65.03% | — | Imported | 2026-05-27 |
| 10 | gemini-1.5-flash-002 (None, important_instructions) | 64.47% | — | Imported | 2026-05-27 |
| 11 | command-r (None, important_instructions) | 63.75% | — | Imported | 2026-05-27 |
| 12 | gpt-4-turbo-2024-04-09 (None, important_instructions) | 62.71% | — | Imported | 2026-05-27 |
| 13 | gpt-3.5-turbo-0125 (None, important_instructions) | 62.16% | — | Imported | 2026-05-27 |
| 14 | claude-3-haiku-20240307 (None, important_instructions) | 62.16% | — | Imported | 2026-05-27 |
| 15 | gpt-4o-2024-05-13 (None, tool_knowledge) | 61.61% | — | Imported | 2026-05-27 |
| 16 | gemini-2.0-flash-exp (None, important_instructions) | 61.44% | — | Imported | 2026-05-27 |
| 17 | gpt-4o-mini-2024-07-18 (None, important_instructions) | 61.37% | — | Imported | 2026-05-27 |
| 18 | gemini-1.5-flash-001 (None, important_instructions) | 60.97% | — | Imported | 2026-05-27 |
| 19 | command-r-plus (None, important_instructions) | 60.34% | — | Imported | 2026-05-27 |
| 20 | gemini-2.0-flash-001 (None, important_instructions) | 59.46% | — | Imported | 2026-05-27 |
| 21 | claude-3-5-sonnet-20240620 (None, important_instructions) | 58.66% | — | Imported | 2026-05-27 |
| 22 | gpt-4o-2024-05-13 (spotlighting_with_delimiting, important_instructions) | 57% | — | Imported | 2026-05-27 |
| 23 | gpt-4o-2024-05-13 (transformers_pi_detector, important_instructions) | 56.59% | — | Imported | 2026-05-27 |
| 24 | claude-3-sonnet-20240229 (None, important_instructions) | 53.26% | — | Imported | 2026-05-27 |
| 25 | gpt-4o-2024-05-13 (None, important_instructions) | 51.2% | — | Imported | 2026-05-27 |
| 26 | gemini-1.5-pro-001 (None, important_instructions) | 50.16% | — | Imported | 2026-05-27 |
| 27 | meta-llama/Llama-3-70b-chat-hf (None, important_instructions) | 46.34% | — | Imported | 2026-05-27 |
| 28 | gpt-4-0125-preview (None, important_instructions) | 42.21% | — | Imported | 2026-05-27 |
No matching rows.