He C.

AI Large Model Training and Optimization Annotator

Text

Participated in the iterative training of a general conversational AI large language model, focusing on dialogue data annotation, evaluation, correction, and fine-tuning. Responsible for cleaning, filtering, and organizing raw dialogue corpora to ensure clean and compliant datasets for training. Conducted manual rating of AI-generated responses and optimized prompts, ensuring logical, fluent, and compliant outputs. • Labeled and rated dialogue data for daily Q&A, life consultation, and generic content scenarios. • Detected errors such as model hallucinations, logic inconsistencies, and knowledge issues, then corrected outputs and refined prompt design. • Reviewed outputs for compliance, filtering sensitive or inappropriate content and refining trigger mechanisms for sensitive words. • Documented daily training issues, measured optimization results, and drafted reports to inform model iteration.