Yue Su

WoG. [ICML 2026] 🏂 World Guidance: World Modeling in Condition Space for Action Generation

DensePolicy. [ICCV 2025] :bouquet: Dense Policy (DSP): Bidirectional Autoregressive Learning of Actions

MBA. [RA-L 2025 & ICRA 2026] :kissing_cat: Motion Before Action: Diffusing Object Motion as Manipulation Condition

DensePolicy2D. [ICCV 2025] :bouquet: 2D version of Dense Policy (DSP)

DSPv2. [ICRA 2026] 🌠 DSPv2: Improved Dense Policy for Effective and Generalizable Whole-body Mobile Manipulation

MetaPalace. Let you in a meta world of The Palace Museum

xdu-thesis-skill. 面向 codex 的西安电子科技大学本科毕业设计（论文）辅助 skill

Selen-Suyue.github.io. Selen's Personal Homepage

AgentCrossTalk. Performe a Crsosstalk by means of two LLM agents

U-pre. U-pre: U-Net is an excellent learner for time series forecasting.

FGSM3D. We designed FGSM adversarial perturbations for point clouds.

ML-Template. If you are also troubled by hand-tearing code, I can provide you with some simple hand-tearing code cases, I hope it can help you

Advpull. Edge feature extraction for ReID and features of generative models. Will continue to be updated later.

CUMCM2023. This is the code we won the provincial first prize in the 2023 National Undergraduate Mathematical Modeling Competition

ML-Homework. My Homerwork at Xidian University 2024 winter concerning ML and Algorithm

Wallpaper. Wallpapers used for building the blog and homepage

biosite. selen's blogs, essays, etc...

DspNet. Project Page of Dense Policy

CUMCM2024. CUMCM 2024 B the code for acquiring the best scheme

UniGen. Unified understanding and generation

Emo_Recognition. Emo Recognition and its pruned model

MBApage. Project page for MBA

BoostMAE. Boost MAE for semantic feature extraction

NP-hard. This is a solution to NP-hard problems represented by TSP. Interestingly, we just chose three optimization algorithms of different dimensions: genetic algorithms represented by traditional GA, evolutionary algorithms represented by DE, and swarm intelligence algorithms represented by TSP.

OpenDoBot. OpenDoBot, a simple visual-motor policy on DoBot Robot.

M-pre. M-pre: Mamba for Time-series Forecasting

Flow_Inpaint. Python

HPE. Human Pose Estimation

ms-bot. multi stage open dobot