SenseNova-U1.5 has been introduced as an 8B native unified multimodal model designed to understand, reason about, and generate visual content within an architecture that omits both encoders and VAEs. According to the paper, this encoder-free and VAE-free design aims to achieve high fidelity and strong instruction following.
The model strengthens its visual interface through spatially coherent patch reconstruction and scales training with carefully curated generation and editing data, improved task formulation, structural prompt enhancement, and native resolutions of up to 4K.
For post-training, the authors optimize specialized experts for visual aesthetics, bilingual text rendering, infographic generation, and image editing, then consolidate their capabilities through multi-expert on-policy distillation.
Evaluations reportedly show that SenseNova-U1.5 advances image fidelity, text rendering, complex composition, multi-reference editing, and interleaved generation, while improving instruction following and preserving subject identity, geometry, and unmodified regions. The paper also notes that despite limited exposure to structured formats in its generation data, the model generalizes effectively to long, complex, and structured visual instructions.
The authors position native unified modelling as a promising path toward systems that perceive, reason, and create within a fully end-to-end framework. They state that training code, including supervised fine-tuning, reinforcement learning, and on-policy distillation, will be open-sourced.