Skip to main content
HomeTopicsModel Robustness

Model Robustness

Cybersecurity statistics about model robustness

Showing 1-9 of 9 results

Multi-turn attack success rate (ASR) ranges from 7.89% to 88.30% across the 15 closed/proprietary flagship models in the cohort.

Cisco5/31/2026
Adversarial AttacksLarge Language Models

Single-turn attack success rate (ASR) ranges from 2.19% to 64.91% across the 15 closed/proprietary flagship models in the cohort.

Cisco5/31/2026
Adversarial AttacksLarge Language Models

Gemini 3 Pro shifts from 18.10% single-turn ASR to 73.35% multi-turn ASR, a 4x increase.

Cisco5/31/2026
Large Language ModelsAdversarial Attacks

Cross-regime deltas (multi-turn ASR minus single-turn ASR) range from −34.74 percentage points to +55.25 percentage points across the cohort.

Cisco5/31/2026
Adversarial AttacksLarge Language Models

Eight of 15 models have an absolute cross-regime gap greater than 15 percentage points.

Cisco5/31/2026
Adversarial AttacksAI Governance

Nova 2 Lite shows 34.05% single-turn ASR but 7.89% multi-turn ASR.

Cisco5/31/2026
Large Language ModelsAdversarial Attacks

Within each multi-turn attack strategy family, the spread between the most- and least-exposed models ranges from 79.51 to 89.25 percentage points.

Cisco5/31/2026
Adversarial AttacksAttack Strategies

Multi-turn attack success rates run 2x to 10x higher than single-turn baselines across eight open-weight LLMs in an earlier evaluation.

Cisco5/31/2026
Adversarial AttacksOpen Models

GPT-5.4 moves from 2.74% single-turn ASR to 24.68% multi-turn ASR, a 9x increase.

Cisco5/31/2026
Large Language ModelsAdversarial Attacks