EasyDeepLearn

What is red-teaming in LLM safety?

medium

Answer

  • Adversarial evaluation: humans and/or other LLMs try to elicit harmful, false, or policy-violating outputs by systematically probing the model.
  • Structured across a taxonomy of harms (hate, violence, sexual content, privacy, CSAM, dangerous instructions, misinformation).
  • Findings feed back into RM data, refusal SFT, and content filters.
  • Anthropic and OpenAI publish red-team reports for major model releases.
  • Automated red-teaming uses attacker LLMs to scale coverage.
Check yourself — multiple choice
  • Only benchmark evaluation
  • Adversarial probing (human + LLM) across a harm taxonomy → feeds back into RM and refusal training
  • Random inputs
  • Only for pretraining

Red-teaming: adversarial probing → feeds back into alignment training.

#safety#red-team#evaluation

Practise LLMs & GenAI

214 interview questions in this topic.

Related questions