Q1. What is the approach to ensure the safety of the gpt-oss models? 🛡️
- Adversarial training
- Neglecting harmful content
- Increasing parameters
- Ignoring user feedback
Answer: Adversarial training
Q2. How do gpt-oss models deal with biases and disallowed content? ⚖️
- They only restrict violent content
- They refuse on a wide range of harmful prompts
- They allow all types of content
- They do not have any safety features
Answer: They refuse on a wide range of harmful prompts
Q3. What evaluation measures the models' performance in detecting harmful prompts? 🔍
- HealthBench
- BBQ evaluation
- AIME
- SWE-Bench Verified
Answer: BBQ evaluation

Test your knowledge of model safety and bias

Related Content From The Pandipedia