Independent Research Project: AI model evaluation, prompt design, and technical benchmarking (3DGS avatar research)
Conducted comparative evaluation and technical benchmarking of frontier LLMs to assess reasoning quality, coding support, and research usefulness. Designed prompts and built AI-assisted research workflows to support paper-writing and agent-based productivity. Performed ongoing model and workflow testing as part of independent research development. • Compared multiple LLMs including GPT-5/5.5, Claude 4/4.7 Opus, DeepSeek, MiniMax, GLM, and Qwen. • Created and iterated prompt designs for structured research assistance. • Used AI-assisted coding and development tools to accelerate debugging and refactoring. • Supported conference/paper outputs and technical decision-making via model evaluation.