Document and Text Task
The scope of my project with Outlier involved prompt writing, AI response evaluation, and data labeling to improve large language model performance. The work focused on creating and refining prompts, assessing model outputs for accuracy and coherence, and ensuring alignment with detailed task guidelines. Projects were typically ongoing and scaled based on client needs, with multiple contributors working under structured workflows. The specific labeling tasks included ranking AI responses, identifying hallucinations or factual errors, evaluating instruction adherence, tagging linguistic issues, and providing written justifications for quality ratings. These tasks required strong analytical skills, attention to nuance, and consistency in applying rubrics across varied content types. Quality measures were strict and performance-driven. All submissions were reviewed or audited, scored against predefined rubrics, and monitored through quality metrics. Maintaining high accuracy, follow