2026 2

Figure from JUICE: Unified Multimodal Modeling in Pretrained Representation Spaces

JUICE: Unified Multimodal Modeling in Pretrained Representation Spaces

Jiahan Li, Huangyuan Su, Shengbang Tong, Shusheng Yang, Ruihan Guo, Suozhi Huang, Hexu Zhao, Pratyusha Sharma, Sherry Yang, Jiatao Gu, Rob Fergus, Kyunghyun Cho

arXiv preprint 2026

One flow matching Transformer, trained from scratch, models images and text jointly in the continuous spaces of frozen pretrained encoders. Text-to-image generation, visual understanding and joint generation become paths through one square of noise levels.

2025 4

2024 5

Figure from Geometric Point Attention Transformer for 3D Shape Reassembly

Geometric Point Attention Transformer for 3D Shape Reassembly

Jiahan Li*, Chaoran Cheng*, Jianzhu Ma, Ge Liu (* equal contribution)

Under Review.

Solving 3D part assembly with the use of Invariant Geometric Point Attention and Pose Recycling.

Figure from Categorical Flow Matching on Statistical Manifolds

Categorical Flow Matching on Statistical Manifolds

Chaoran Cheng*, Jiahan Li*, Jian Peng, Ge Liu (* equal contribution)

Conference and Workshop on Neural Information Processing Systems (NeurIPS) 2024

A novel flow matching framework on the manifold of parameterized probability measures for discrete generation.

2022 2

Figure from Proximal Exploration for Model-guided Protein Sequence Design

Proximal Exploration for Model-guided Protein Sequence Design

Zhizhou Ren*, Jiahan Li*, Fan Ding, Yuan Zhou, Jianzhu Ma, Jian Peng (* equal contribution)

International Conference on Machine Learning (ICML) 2022 Spotlight

Addressing model-guided sequence design using evolutionary search to find high-fitness mutants with low mutation counts.