Bilingual AI Content Evaluator
As a Bilingual AI Content Evaluator, I conducted RLHF (Reinforcement Learning from Human Feedback) by ranking and labeling LLM responses for helpfulness, honesty, and harmlessness. My work included advanced linguistic evaluations for Cantonese and English code-switching, focusing on cultural and grammatical appropriateness in the Hong Kong context. The job also required detailed technical justification of model rankings and logic pathways. • Labeled text data from large language models, identifying and mitigating hallucinations. • Conducted fact-checking and validation to ensure response accuracy. • Ensured high-quality linguistic and cultural evaluations for both Cantonese and English. • Drafted reasoned technical summaries in English to support model feedback decisions.