Yongxian Wei

MLLMerging. ICLR 2026 "OptMerge: Unifying Multimodal LLM Capabilities and Modalities via Model Merging".

DOGE. Official implementation of "Modeling Multi-Task Model Merging as Adaptive Projective Gradient Descent".

CVLM. Official implementation of "Open-Vocabulary Customization from CLIP via Data-Free Knowledge Distillation".

MLWDUDM. Official implementation of "Meta-Learning Without Data via Unconditional Diffusion Model".

FREE. Official implementation of "FREE: Faster and Better Data-Free Meta-Learning".

TGR. Official implementation of "Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models".

Reasoning-Synthesis. Learning to Pose Problems: Reasoning‑Driven and Solver‑Adaptive Data Synthesis