BOLD
Gemma 7B 49.1 Only in system cards Models continue 23,679 English Wikipedia prompts about professions, gender, race, religion and politics. Classifiers score sentiment, toxicity and polarity per group.
Results are given per group, usually as mean and spread of sentiment.
In the authors’ words
Recent advances in deep learning techniques have enabled machines to generate cohesive open-ended text when prompted with a sequence of words as context. While these models now empower many downstream applications from conversation bots to automatic storytelling, they have been shown to generate texts that exhibit social biases. To systematically study and benchmark social biases in open-ended language generation, we introduce the Bias in Open-Ended Language Generation Dataset (BOLD), a large-scale dataset that consists of 23,679 English text generation prompts for bias benchmarking across five domains: profession, gender, race, religion, and political ideology. We also propose new automated metrics for toxicity, psycholinguistic norms, and text gender polarity to measure social biases in open-ended text generation from multiple angles. An examination of text generated from three popular language models reveals that the majority of these models exhibit a larger social bias than human-written Wikipedia text across all domains. With these results we highlight the need to benchmark biases in open-ended language generation and caution users of language generation models on downstream tasks to be cognizant of these embedded prejudices.
- Level
- Everyday Most adults, without training
- Scored
- Graded by a model A language model scores each response, usually against a rubric.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Gemma 7B leads with 49.1
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.
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.
7 system cards report it, from 3 labs
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 |
|---|---|---|---|---|
| 13 Mar 2024 | Gemma 7B technical report | 49.1 | BOLD, IT (Table 8) | Bias |
| 13 Mar 2024 | Gemma 2B technical report | 45.6 | BOLD, IT (Table 8) | Bias |
| 8 Jan 2024 |  Mixtral 8x7B technical report | BOLD sentiment score (avg ± std) for gender, profession, religious_ideology, political_ideology, race; no overall (Figure 5 table) | Bias | |
| 19 Jul 2023 |  Llama 2-7B technical report | BOLD mean sentiment per demographic group across race, gender, religion, politics and profession, pretrained and Llama 2-Chat, no overall (Tables 46 to 50) | Bias | |
| 19 Jul 2023 | Llama 2-70B technical report | BOLD mean sentiment per demographic group across race, gender, religion, politics and profession, pretrained and Llama 2-Chat, no overall (Tables 46 to 50) | Bias | |
| 19 Jul 2023 | Llama 2-34B technical report | BOLD mean sentiment per demographic group across race, gender, religion, politics and profession, pretrained and Llama 2-Chat, no overall (Tables 46 to 50) | Bias | |
| 19 Jul 2023 | Llama 2-13B technical report | BOLD mean sentiment per demographic group across race, gender, religion, politics and profession, pretrained and Llama 2-Chat, no overall (Tables 46 to 50) | 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%