AttaQ
Models answer 1,402 adversarial questions in seven harm classes, such as deception, violence and PII. Responses are scored for harmlessness.
Questions were extracted from Anthropic hh-rlhf, generated by LLMs, or built from a Wikipedia list of crimes. MIT license.
- Made by
- IBM Research
- Released
- September 2023
- Kind
- Refusing harmful requests
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Last in a launch post
- April 2026, dropped after it saturated
- Last model added
- April 2026
In the authors’ words
As large language models become more prevalent, their possible harmful or inappropriate responses are a cause for concern. This paper introduces a unique dataset containing adversarial examples in the form of questions, which we call AttaQ, designed to provoke such harmful or inappropriate responses. We assess the efficacy of our dataset by analyzing the vulnerabilities of various models when subjected to it. Additionally, we introduce a novel automatic approach for identifying and naming vulnerable semantic regions - input semantic areas for which the model is likely to produce harmful outputs. This is achieved through the application of specialized clustering techniques that consider both the semantic similarity of the input attacks and the harmfulness of the model's responses. Automatically identifying vulnerable semantic regions enhances the evaluation of model weaknesses, facilitating targeted improvements to its safety mechanisms and overall reliability.
Granite 3.3 8B Instruct leads with 88.5% among models released by April 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in April 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 Granite 3.3 8B Instruct 88.5%
- 2 Granite 4.0 H Small 86.6%
- 3 Granite 4.1 30B 85.8%
- 4 Granite 3.2 8B Instruct 85.4%
- 5 granite-4.0-1b 85.3%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Granite 3.3 8B Instruct | 88.5% | Apr 2025 | Lab model card | |
| 2 | Granite 4.0 H Small | 86.6% | Oct 2025 | Lab model card | |
| 3 | Granite 4.1 30B | 85.8% | Apr 2026 | Lab model card | |
| 4 | Granite 3.2 8B Instruct | 85.4% | Feb 2025 | Lab model card | |
| 5 | granite-4.0-1b | 85.3% | Oct 2025 | Lab model card |
2 models have held number one in labs’ own numbers since its release in September 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
IBM’s own number for Granite 3.2 8B Instruct passed 85% in February 2025
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs headline it less often now, last in April 2026
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Apr 29, 2026 | Granite 4.1 30B | 85.8 | AttaQ |
| Oct 28, 2025 | granite-4.0-1b | 85.3 | AttaQ |
| Oct 2, 2025 | Granite 4.0 H Small | 86.6 | AttaQ |
| Apr 16, 2025 | Granite 3.3 8B Instruct | 88.5 | AttaQ |
| Feb 26, 2025 | Granite 3.2 8B Instruct | 85.4 | AttaQ |
More in Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6% - Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
- DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics.
Muse Spark 1.2 88.4% - Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
- IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%