AI/LLM Analyst
Internal Proprietary ToolingImageBounding BoxClassification
The project focused on evaluating and comparing AI-generated responses to image-based prompts in a Vision Text-to-Image Head-to-Head (TI2T H2H) setting. The primary data labeling tasks included categorizing prompts, rating responses across dimensions such as factuality, instruction following, helpfulness, and style/format, and selecting the stronger response with justification. The project involved reviewing a large set of image-text pairs across diverse domains including OCR, visual reasoning, document analysis, and general knowledge. Quality measures adhered to strict evaluation guidelines emphasizing accuracy, completeness, consistency, adherence to prompt instructions, and objective comparative judgment.
The project focused on evaluating and comparing AI-generated responses to image-based prompts in a Vision Text-to-Image Head-to-Head (TI2T H2H) setting. The primary data labeling tasks included categorizing prompts, rating responses across dimensions such as factuality, instruction following, helpfulness, and style/format, and selecting the stronger response with justification. The project involved reviewing a large set of image-text pairs across diverse domains including OCR, visual reasoning, document analysis, and general knowledge. Quality measures adhered to strict evaluation guidelines emphasizing accuracy, completeness, consistency, adherence to prompt instructions, and objective comparative judgment.
2025