Baidu ERNIE team unveils HiUni: one dense backbone unifies visual understanding and generation
ernie-research · hf · 2026-11-20
Baidu's ERNIE team presents HiUni, extending a pretrained VLM into a unified multimodal model with heavy parameter sharing: the same language-aligned ViT and dense Transformer backbone handles both understanding and generation, with task-specific computation confined to the output side. A Pixel Diffusion Head tackles pixel-space optimization, and a lightweight HiGate pathway restores fine spatial details lost to ViT abstraction. HiUni matches the base VLM on understanding while staying competitive on image generation and editing.
More from Multimodal
- fofrAI experiments with Neon and looping GLSL shaders for generative visuals — fofrAI · 2026-11-20
- LiliPixel v1: a from-scratch 190M pixel-art DiT under a fully open license — ApprehensiveTart3158 · 2026-11-20
- Seller builds ComfyUI 3x3 grid workflow turning phone snaps into product shots at $0.0355/image — Niuaia · 2026-11-20
- Astra 6.2 achieves unprecedented 3D realism in image generation with extra steering — gabrielchua · 2026-11-20
- Nano Banana Pro 2 staged for release at Google, leak claims — mark_k · 2026-11-20
- 开源 MCP 服务器让 Claude 写 Strudel 代码现场作曲并驱动真实 MIDI 合成器 — konni_away · 2026-11-20