Language Systems
We focus on prompting, retrieval storytelling, and conversational structure so model capability can enter real workflows.
A research-driven AI prototype lab
Turning research questions into products people can actually experience
Dianzinao Lab explores language, vision, and agent systems, then turns abstract methods into interactive prototypes that can be demonstrated, evaluated, and iterated quickly.
Research Highlights
Three tracks, organized around one goal
We do not present isolated technical topics. We keep asking which capabilities deserve to become real, usable, and testable product experiences.
Vision Understanding
From interface recognition to information extraction, we test how visual models behave inside real tasks.
Agent Orchestration
We explore task decomposition, tool use, and state feedback to build more reliable automation prototypes.
Latest Research
DATECATEGORYTITLE