Stars
We propose Reinforcement Learning from Community Feedback (RLCF), a training paradigm that uses large-scale community signals as supervision, and formulate scientific taste learning as a preference…
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
🔥 LeetCode for PyTorch — practice implementing softmax, attention, GPT-2 and more from scratch with instant auto-grading. Jupyter-based, self-hosted or try online.
slime is an LLM post-training framework for RL Scaling.
The CS61A course of UC Berkeley. A python version of SICP.
A basic framework for testing everything in a maching learning model.
Official implementation of FedSub: An efficient subspace algorithm for federated learning on heterogeneous data.
Lightweight data loader with zero-padding sentence packing for LLM training.
Training library for Megatron-based models with bidirectional Hugging Face conversion capability
EasyR1: An Efficient, Scalable, Multi-Modality RL Training Framework based on veRL
[CVM'26]BoCoR-Seg: Bidirectional Co-Refinement Framework for High-Resolution Remote Sensing Image Segmentation
The code of Paper "Locate Then Ask: Interpretable Stepwise Reasoning for Multi-hop Question Answering".
The source code of Paper "PathQG: Neural Question Generation from Facts".
The code of Paper "Logic-Driven Context Extension and Data Augmentation for Logical Reasoning of Text".
Implementation of "Interleaved Latent Visual Reasoning with Selective Perceptual Modeling".
[NeurIPS 2025 Spotlight] SparseMVC: Probing Cross-view Sparsity Variations for Multi-view Clustering [Pytorch repository]
[CVPR 2026] Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens
https://huggingface.co/datasets/multimodal-reasoning-lab/Zebra-CoT
Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.
A cross-platform bilibili toolbox. 跨平台哔哩哔哩工具箱，支持下载视频、番剧等等各类资源
An ASR (Automatic Speech Recognition) adversarial attack repository.
🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
Enhanced Deep Image Prior for Unsupervised Hyperspectral Image Super-resolution, TGRS. (PyTorch)
A repo recording: How I use UNet to solve problems
🚀 Cross attention map tools for huggingface/diffusers