The best ChatGPT that $100 can buy.

Pytorch code for NeurIPS 2025 paper "Accurate and Efficient Low-Rank Model Merging in Core Space"

[ICML 2025] No Task Left Behind: Isotropic Model Merging with Common and Task-Specific Subspaces (official repository)

A Django web application to track 2v2 (spikeball) matches internally in my lab

AlphaFold 3 inference pipeline.

Open source code for AlphaFold 2.

Code for paper "Boosting Continual Learning of Vision-Language Models via Mixture-of-Experts Adapters" CVPR2024

Official implementation of the NeurIPS 2025 paper "Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space"

[NeurIPS 2022] “M³ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-design”, Hanxue Liang*, Zhiwen Fan*, Rishov Sarkar, Ziyu Jiang, Tianlong Che…

Combating hidden stratification with GEORGE

[WACV 2025] Debiasify

Official implementation of BPA (CVPR 2022)

Robust fine-tuning of zero-shot models

An open-source framework for training large multimodal models.

An open source implementation of CLIP.

Pre-trained models, data, code & materials from the paper "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness" (ICLR 2019 Oral)

[ICLR 2025] Official implementation of "Mitigating Parameter Interference in Model Merging via Sharpness-Aware Fine-Tuning"

Biomni: a general-purpose biomedical AI agent

Repository for the paper "U-Net Transplant: The Role of Pre-training for Model Merging in 3D Medical Segmentation" accepted @ MICCAI2025

Official codebase of "Update Your Transformer to the Latest Release: Re-Basin of Task Vectors" - ICML 2025

Task Singular Vectors: Reducing Task Interference in Model Merging. Merge models avoiding task interference through separable models.

Pruna is a model optimization framework built for developers, enabling you to deliver faster, more efficient models with minimal overhead.

[NeurIPS'24] Official PyTorch implementation for paper "Knowledge Composition using Task Vectors with Learned Anisotropic Scaling"

Codebase for "C2M3: Cycle-Consistent Multi-Model Merging".

A python library for self-supervised learning on images.

Patching open-vocabulary models by interpolating weights

A framework for merging models solving different tasks with different initializations into one multi-task model without any additional training

Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time