CrowS-Pairs
Gemma 7B 51.3% leads since Mar 2024 Only in system cards Models choose between paired sentences that differ only in a stereotyped social group. Measures preference for the stereotyped sentence.
A score of 50 percent means no preference between stereotyped and anti-stereotyped sentences.
In the authors’ words
Pretrained language models, especially masked language models (MLMs) have seen success across many NLP tasks. However, there is ample evidence that they use the cultural biases that are undoubtedly present in the corpora they are trained on, implicitly creating harm with biased representations. To measure some forms of social bias in language models against protected demographic groups in the US, we introduce the Crowdsourced Stereotype Pairs benchmark (CrowS-Pairs). CrowS-Pairs has 1508 examples that cover stereotypes dealing with nine types of bias, like race, religion, and age. In CrowS-Pairs a model is presented with two sentences: one that is more stereotyping and another that is less stereotyping. The data focuses on stereotypes about historically disadvantaged groups and contrasts them with advantaged groups. We find that all three of the widely-used MLMs we evaluate substantially favor sentences that express stereotypes in every category in CrowS-Pairs. As work on building less biased models advances, this dataset can be used as a benchmark to evaluate progress.
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Multiple choice
Gemma 7B leads with 51.3% among models released by Jun 2024
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 Jun 2024, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 Gemma 7B 51.3%
- 2 Gemma 2B 45.8%
- 3 Gemma 2 9B 37.5%
- 4 Gemma 2 27B 36.7%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemma 7B | 51.3% | Mar 2024 | Model card | |
| 2 | Gemma 2B | 45.8% | Mar 2024 | Model card | |
| 3 | Gemma 2 9B | 37.5% | Jul 2024 | Model card | |
| 4 | Gemma 2 27B | 36.7% | Jul 2024 | Model card |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
4 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 31 Jul 2024 | Gemma 2 9B technical report | 37.5 | CrowS-Pairs, top-1, IT (Table 18) | Bias |
| 31 Jul 2024 | Gemma 2 27B technical report | 36.7 | CrowS-Pairs, top-1, IT (Table 18) | Bias |
| 13 Mar 2024 | Gemma 7B technical report | 51.3 | CrowS-Pairs, top-1, IT (Table 8) | Bias |
| 13 Mar 2024 | Gemma 2B technical report | 45.8 | CrowS-Pairs, top-1, IT (Table 8) | Bias |
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% - AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
- MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy.
Muse Spark Open 98.1% - 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. GPT-6 Astra 8.50%
- Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%