LessonPlanner embeds Gagné’s Nine Events of Instruction into an LLM-driven lesson planning interface, guiding novice teachers to interactively construct pedagogically sound lesson plans.

LitLinker is an LLM-powered interactive system that helps elementary school literature teachers ideate interdisciplinary contexts by linking reading materials to subjects such as science and art.

The study reveals that GenAI (deployed as AskArt) benefits students by providing background information, creative inspirations, and personalized guidance, while also identifying challenges such as difficulty in query formulation, varied student collaboration strategies, and teacher concerns about misuse.

ReDBot is an LLM-powered conversational recommendation bot for group chats that proactively asks questions to identify group preferences and recommends alternatives accordingly to facilitate consensus in online group decision-making tasks.

By extracting low-level texture priors via Canny edge detection and fusing them with multi-scale deep features, TPMN addresses the challenges of non-uniform surface textures and multi-scale defect openings in logistics parcel inspection, achieving strong sensitivity to defective packages on a self-collected cardboard-boxes dataset.

AtomThink introduces four key modules — a data engine, atomic step fine-tuning, policy-guided multi-turn inference, and an atomic capability metric — to bring structured slow-thinking capabilities into visual understanding models, while improving data utilization by 5× and inference efficiency by 85.3% over state-of-the-art structured CoT methods.