The Prism Hypothesis unified autoencoding paper accepted at ECCV 2026, paves way for encoder-free MLLMs

liuziwei7 · x · 2026-09-10

"The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding" was accepted to ECCV 2026 (poster, Sep 10, Malmö). It finds image generation and understanding components share the same low-frequency space, enabling an encoder-free unified MLLM (SenseNova-U1); a "v2" on pixel-space diffusion also shipped. Code and paper are public.

Original post →

More from Multimodal

Multimodal channel →