Domain Maps

认知地图

Domain Maps 是研究方向的认知结构图，帮助建立概念框架和追踪知识演进。

核心研究方向

当前聚焦六大领域：

1. GUI Agent — Grounding Robustness / RL Training / Self-Improving
2. Agentic RL — GRPO / Credit Assignment / Self-Improving / Test-Time RL
3. VLM — 高分辨率编码 / 理解-生成统一 / 效率优化 / Alignment
4. Embodied AI — VLA Foundation / Diffusion Policy / World Model / Safety
5. World Model — Video World Model / DGE / Environment Synthesis / UI World Model
6. Hyperbolic Manifold — Hyperbolic Embeddings / GNN / Attention / Contrastive Learning / Scalable KG / Lorentz GNN / Riemannian Optimization