GeezerVison

This is the 360 AI Research, our mission is to lead in tech innovations and deliver real-world values.
We focus on "multimodal + cross-modal learning" and "large model + zero/few shot learning",
conducting research in

• 🔎 multi-modal comprehension

  • FG-CLIP: ICML2026 and ICML2025, new generation of CLIP with strong fine grained discrimination capability
  • RzenEmbed: Embedding model prioritized towards Multimodal RAG, overall + VisDoc double top1 on MMEB benchmark
  • GAG: Generation Augmented Generation, resolving fragmented knowledge defect of RAG
  • LMM-Det: ICCV2025, make large multimodal models excel in object detection
  • IAA: AAAI2025, LMM with plugin mechanism solving catastrophic forgetting
  • 360VL: Large multimodal model, 2nd-gen
  • SEEChat: Multimodal chatbot with computer vision capabilities integrated, our 1st-gen LMM
  • OVD: KDD2023, open-world object detection, we also co-hosted open vocabulary detection contest 2023 with CSIG(中国图象图形学学会)
  • Zero: ACM MM2023, large scale open-sourced Chinese cross-modal data and benchmark

• 🎨 multi-modal generation

  • RevealLayer: ICML2026, CONTROLLABLE layer-decomposition/object-wise editing/matting
  • NAMI-T2I: CVPR2026, efficient DiT architecture for T2I
  • FLUX-Makeup: Transfer face makeup with high fidelity, SoTA and production-ready
  • RefTon: CVPR2026, Reference person shot assist virtual Try-On
  • PlanGen: ICCV2025, unified layout planning and image generation
  • Qihoo-T2X: ICLR2025, efficient DiT architecture for text2any tasks
  • BDM: AAAI2025, Chinese-native image generation while compatible with SD eco-system, 1st-gen
  • HiCo: NeurIPS2024, layout controlled image generation
  • FancyVideo: Video generation from text&image, 1st-gen

Check research.360.cn for contact and API portal

Internship: we're hiring research interns in fileds of AIGC, LMM, and inference optimization, check 👉 JD here