Evaluate DALL-E 2 diversity improvements with user feedback
Assess DALL-E 2's image diversity by asking users if generated people reflect global diversity.
Ask users if DALL-E images included people of diverse backgrounds after a new technique was applied.
Users were 12x more likely to say DALL-E images included people of diverse backgrounds after the technique was applied.
After implementing the diversity technique, users reported a 12x increase in perceived diversity in generated images of people.
Evaluate the effectiveness of DALL-E 2's new diversity technique by measuring user perception. The goal is to confirm if generated images of people, when prompts don't specify race or gender, more accurately reflect global diversity.
User feedback directly measures the perceived impact of the system-level technique on image diversity, providing a clear metric for improvement.
- Reducing bias and improving safety in DALL·E 2 ↗OpenAI · OpenAI News · Jul 18, 2022 · paraphrased
First seen on Subvolts Jul 18, 2022. Updated Sep 11, 2026. Spotted a problem? Tell us.
- Reducing bias and improving safety in DALL·E 2OpenAI · OpenAI News · Jul 18, 2022
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- Case studyUsing Screwtape to find confirmed failures of AI monitorsDemonstrates how a red-teamer can find confirmed failures of an AI monitor without developer infrastructure access. · 120 min
- ChecklistCheck AI model safety evidence acquisition policiesEvaluate how AI models seek safety information before acting using a controlled benchmark. · 120 min
- RuleVerify AI-Generated Biological Designs with WatermarksEmbed watermarks in AI-generated biological designs to ensure authenticity and prevent misuse. · 60 min
- ChecklistEvaluating LLM N-day exploit generationSteps to assess an LLM's ability to turn security patches into proof-of-concept crashes. · 30 min
- Eval methodMethod for evaluating potential discriminatory impact of LMsUse an LM to generate prompts across scenarios, varying demographics to find discrimination. · 60 min
- HeuristicEvolve multi-turn LLM attacks using phased conversation plansTreat attack strategies as phased conversation plans, not raw prompts, for LLM red-teaming. · 30 min
Questions people ask
- Which AI tools does this work with?
- Any current model. It was seen with in the sources.
- Where did this come from?
- OpenAI (OpenAI News). Subvolts collected, normalized and credited it; the link above opens the original.