AI Text Annotation & Quality Evaluation for LLM Training
Worked on text-based data labeling and evaluation tasks focused on improving Large Language Model (LLM) outputs. The project involved reviewing, annotating, and rating AI-generated responses based on predefined guidelines for accuracy, relevance, tone, completeness, and instruction-following. Key tasks included: Evaluating AI-generated text outputs and assigning quality ratings Comparing multiple model responses and selecting the most appropriate output Annotating summaries for correctness and information coverage Classifying text data into relevant categories Writing and refining prompts and ideal responses for supervised fine-tuning (SFT) Identifying errors, inconsistencies, and potential bias in AI-generated content Maintained high accuracy and consistency by strictly following annotation guidelines, performing self-quality checks, and documenting edge cases. Delivered clean, well-structured labeled data suitable for model training and improvement.