LLM Training & Response Evaluation Project
Assisted in a large-scale LLM training and evaluation task with a focus on enhancing the overall performance of conversational AI models. This entailed annotating text data for intent detection, sentiment analysis, and named entity recognition (NER) categories. It also involved crafting high-quality prompt-response pairs for SFT. Assessed the relevance, accuracy, safety, tone, and overall compliance with specific guidelines of AI-generated text within the scope of RLHF. This task comprised labeling thousands of data points with stringent quality requirements, multiple levels of reviews, and constant calibration.