AI Evaluator—LLM Output Evaluation and Safety Testing
TextI participated in the evaluation of large language model (LLM) outputs for correctness, safety, and compliance with guidelines. My work involved ranking, reviewing, and annotating model responses in agentic task workflows. I focused on adversarial prompt design and model output assessment to support AI safety measures. • Reviewed and rated model outputs for factual accuracy and appropriateness. • Conducted RLHF annotation tasks on generated text to fine-tune LLM performance. • Designed and tested adversarial prompts to evaluate model resilience to unsafe behaviors. • Collaborated with AI researchers to assess and improve agentic workflows.
2024 - Present