Code release for UNCHA: UNcertainty-guided Compositional Hyperbolic Alignment (CVPR 2026 Highlight)

Solving Token Gradient Conflict in Mixture-of-Experts for Large Vision-Language Model

GRadient-INformed MoE

[ICML 2025 Oral] Mixture of Lookup Experts

Code for paper 'How does Transformer Learn Implicit Reasoning?'

The code of Advancing Expert Specialization for Better MoE (NeurIPS2025 oral)

Cosmos-Predict1 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

PyTorch implementation of JiT https://arxiv.org/abs/2511.13720

Pytorch Implementation (unofficial) of the paper "Mean Flows for One-step Generative Modeling" by Geng et al.

Code for ICLR 2025 Paper "Projection Head is Secretly an Information Bottleneck"

Implementation of Multi-View Information Bottleneck

RoFormer V1 & V2 pytorch

[ECCV 2024] Official PyTorch implementation of RoPE-ViT "Rotary Position Embedding for Vision Transformer"

Rotary Transformer

[NeurIPS 2024 Spotlight] Code for the paper "Flex-MoE: Modeling Arbitrary Modality Combination via the Flexible Mixture-of-Experts"

Astrofy is a free and open-source template for your Personal Portfolio Website built with Astro and TailwindCSS. Create in minutes a website with Blog, CV, Project Section, Store and RSS Feed.

A beautiful, simple, clean, and responsive Jekyll theme for academics

Implementation of the "the first large-scale multimodal mixture of experts models." from the paper: "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts"

Keras implementation of Representation Learning with Contrastive Predictive Coding

[NeurIPS 2021] Multiscale Benchmarks for Multimodal Representation Learning

Code for PHATGOOSE introduced in "Learning to Route Among Specialized Experts for Zero-Shot Generalization"

AI notepad for meetings

[NeurIPS 2023] Factorized Contrastive Learning: Going Beyond Multi-view Redundancy