Changjiang Gao

TESSY. A Teacher–Student Cooperation Framework to Synthesize Student-Consistent SFT Data

mempalace. The best-benchmarked open-source AI memory system. And it's free.

reverse-engineering-gemma-3n. Reverse Engineering Gemma 3n: Google's New Edge-Optimized Language Model

CLI-Anything. "CLI-Anything: Making ALL Software Agent-Native" -- CLI-Hub: https://clianything.cc/

BLEnD. BLEnD: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages

spider. Streamline on-policy/off-policy distillation workflows in a few lines of code

trl. Train transformer language models with reinforcement learning.

Math-Verify. Python

Develpoment-of-Upper-Limb-Exoskeleton. Upper Limb Exoskeleton development

LEAP-upper-limb. Upper limb powered exoskeleton (Stop updating)

de_vito. Kinematic control algorithms and software for 🦾 DE VITO 🦾: A dual-arm, high degree-of-freedom, lightweight, inexpensive, passive upper-limb exoskeleton for robot teleoperation.

dataset-ssvep-exoskeleton. SSVEP-based BCI recording of 12 subjects operating an upper limb exoskeleton during a shared control task. The exoskeleton is either controlled with a touchless interface detecting hand poses or with BCI.

VST. [ECCV2026] Visual Spatial Tuning

nanochat. The best ChatGPT that $100 can buy.

SAELens. Training Sparse Autoencoders on Language Models

JanusCoder. [ICLR 2026] JanusCoder: Towards a Foundational Visual-Programmatic Interface for Code Intelligence

Embodied-AI-Guide. [Lumina具身智能社区] 具身智能技术指南 Embodied-AI-Guide

algonauts-2025. Training and evaluating encoding models to predict fMRI brain responses to naturalistic video stimuli

unsloth. Unsloth is a local UI for training and running Kimi K3, Gemma 4, Qwen3.6, DeepSeek, GLM and other models.

ms-swift. Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).

LoRA-Models-for-SAEs. Code for reproducing our paper "Low Rank Adapting Models for Sparse Autoencoder Features"

Gradient_Unified. How Instruction and Reasoning Data shape Post-Training: Data Quality through the Lens of Layer-wise Gradients

circuit-tracer. Python

mtob. Shell

MUR. Python

OpenAI_WebUI. WebUI to various OpenAI API-compatible GPTs and Image generation API (requires valid API keys)

open-thoughts. Fully open data curation for reasoning models

uv. An extremely fast Python package and project manager, written in Rust.

kodcode. ✨ A synthetic dataset generation framework that produces diverse coding questions and verifiable solutions - all in one framwork

CapArena. An Arena-style Automated Evaluation Benchmark for Detailed Captioning

DeepMath. A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning

TLDR. Code for Research Project TLDR

s1. s1: Simple test-time scaling

Logic-LLM. The project page for "LOGIC-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning"

ScienceBoard. [ICLR 2026] Code, benchmark and environment for "ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows"

rllm. Democratizing Reinforcement Learning for LLMs

Absolute-Zero-Reasoner. Official Repository of Absolute Zero Reasoner

dingo. Dingo: A Comprehensive AI Data, Model and Application Quality Evaluation Tool

QuRating. [ICML 2024] Selecting High-Quality Data for Training Language Models

continuous-thought-machines. Continuous Thought Machines, because thought takes time and reasoning is a process.

atom. [NeurIPS 2025] Atom of Thoughts for Markov LLM Test-Time Scaling

TFB. [PVLDB 2024 Best Paper Nomination] TFB: Towards Comprehensive and Fair Benchmarking of Time Series Forecasting Methods

rbloom. A fast, simple and lightweight Bloom filter library for Python, implemented in Rust.

verl. verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

SuperGPQA. Python

steplaw. Python

MAPO. The implement of ACL2024: "MAPO: Advancing Multilingual Reasoning through Multilingual Alignment-as-Preference Optimization"

near-duplicate-code-detector. A simple tool for detecting near-duplicate source code

BenchMAX. Python

open-r1. Fully open reproduction of DeepSeek-R1

llm-reasoners. A library for advanced large language model reasoning

SimPO. [NeurIPS 2024] SimPO: Simple Preference Optimization with a Reference-Free Reward

chat_templates. Chat Templates for 🤗 HuggingFace Large Language Models

pydec. Linear decomposition toolkit for neural network.

xtuner. A Next-Generation Training Engine Built for Ultra-Large MoE Models

LRP-eXplains-Transformers. Layer-wise Relevance Propagation for Large Language Models and Vision Transformers [ICML 2024]

captum. Model interpretability and understanding for PyTorch

Open-LLaVA-NeXT. An open-source implementation for training LLaVA-NeXT.

sglang. SGLang is a high-performance serving framework for large language models and multimodal models.

human-eval. Code for the paper "Evaluating Large Language Models Trained on Code"

lingua-py. The most accurate natural language detection library for Python, suitable for short text and mixed-language text

lm-evaluation-harness. A framework for few-shot evaluation of language models.

dedupe. :id: A python library for accurate and scalable fuzzy matching, record deduplication and entity-resolution.

deduplicate-text-datasets. Rust

MMBench. Official Repo of "MMBench: Is Your Multi-modal Model an All-around Player?"

lmms-eval. One-for-All Multimodal Evaluation Toolkit Across Text, Image, Video, and Audio Tasks

LLaVA-NeXT. Python

belebele. Repo for the Belebele dataset, a massively multilingual reading comprehension dataset.

mt-bench-101. [ACL 2024] MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialogues

Deep-Fourier-based-Arbitrary-scale-Super-resolution-for-Real-time-Rendering. SIGGRAPH 2024 Conference Paper: Deep Fourier-based Arbitrary-scale Super-resolution for Real-time Rendering

InternVL. [CVPR 2024 Oral] InternVL Family: A Pioneering Open-Source Alternative to GPT-4o. 接近GPT-4o表现的开源多模态对话模型

VLMEvalKit. Open-source evaluation toolkit of large multi-modality models (LMMs), support 220+ LMMs, 80+ benchmarks

awesome-brain-decoding. collection of awesome research in brain decoding, including interaction with multi-modalities, theories, and foundation models.

the-story-of-heads. This is a repository with the code for the ACL 2019 paper "Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned" and the ACL 2021 paper "Analyzing Source and Target Contributions to NMT Predictions".

pydec. Linear decomposition toolkit for neural network.

Classical-Modern. 非常全的文言文（古文）-现代文平行语料

qlora. QLoRA: Efficient Finetuning of Quantized LLMs

gfn-lm-tuning. Jupyter Notebook

LlamaFactory. Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)

LWM. Large World Model -- Modeling Text and Video with Millions Context

vllm. A high-throughput and memory-efficient inference and serving engine for LLMs

EasyNLP. EasyNLP: A Comprehensive and Easy-to-use NLP Toolkit

XGLUE. Cross-lingual GLUE

LLMs-are-parallel-multilingual-learners. The implementation of Revealing the Parallel Multilingual Learning within Large Language Models (EMNLP Main 2024).

MLQA. New dataset

mlmm-evaluation. Multilingual Large Language Models Evaluation Benchmark

grok-1. Grok open release

Cross-Lingual-MRC. Cross-Lingual Machine Reading Comprehension (EMNLP 2019)

NextChat. ✨ Light and Fast AI Assistant. Support: Web | iOS | MacOS | Android | Linux | Windows

llm-latent-language. Repo accompanying our paper "Do Llamas Work in English? On the Latent Language of Multilingual Transformers".

Prompt-Engineering-Guide. 🐙 Guides, papers, lessons, notebooks and resources for prompt engineering, context engineering, RAG, and AI Agents.

llm-course. Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.

brain_syntactic_representations. Code to produce syntactic representations that can be used to study syntax processing in the human brain

QAlign. Ruby

mLoRA. An Efficient "Factory" to Build Multiple LoRA Adapters

Chinese-LLaMA-Alpaca-2. 中文LLaMA-2 & Alpaca-2大模型二期项目 + 64K超长上下文模型 (Chinese LLaMA-2 & Alpaca-2 LLMs with 64K long context models)

TracIn. Implementation of Estimating Training Data Influence by Tracing Gradient Descent (NeurIPS 2020)

GAOKAO-Bench. GAOKAO-Bench is an evaluation framework that utilizes GAOKAO questions as a dataset to evaluate large language models.

XCSR. Code Repo for the ACL21 paper "Common Sense Beyond English: Evaluating and Improving Multilingual LMs for Commonsense Reasoning"

LAMA. LAnguage Model Analysis

X-FACTR. Python

WhatICLLearns. [ACL 2023 Findings] What In-Context Learning “Learns” In-Context: Disentangling Task Recognition and Task Learning

Chinese-LLaMA-Alpaca. 中文LLaMA&Alpaca大语言模型+本地CPU/GPU训练部署 (Chinese LLaMA & Alpaca LLMs)

instruct-eval. This repository contains code to quantitatively evaluate instruction-tuned models such as Alpaca and Flan-T5 on held-out tasks.

ff-layers. The accompanying code for "Transformer Feed-Forward Layers Are Key-Value Memories". Mor Geva, Roei Schuster, Jonathan Berant, and Omer Levy. EMNLP, 2021.

scibench. Python

relative_importance. Python

ValueZeroing. The official repo for the EACL 2023 paper "Quantifying Context Mixing in Transformers"

attention_flow. Jupyter Notebook

transformer-contributions. Measuring the Mixing of Contextual Information in the Transformer

logit-explanations. Jupyter Notebook

norm-analysis-of-transformer. Jupyter Notebook

cross_lingual_relative_importance. Relative word importance of different languages and multilingualism (new work)

modelblocks-release. A psycholinguistic modeling toolkit

lm-question-generation. Multilingual/multidomain question generation datasets, models, and python library for question generation.

CoGnition. Python

brain-inspired-replay. A brain-inspired version of generative replay for continual learning with deep neural networks (e.g., class-incremental learning on CIFAR-100; PyTorch code).

knn-box. an easy-to-use knn-mt toolkit

test. Measuring Massive Multitask Language Understanding | ICLR 2021

RWKV-LM. RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding.

alpaca-lora. Instruct-tune LLaMA on consumer hardware

LMFlow. An Extensible Toolkit for Finetuning and Inference of Large Foundation Models. Large Models for All.

langchain. The agent engineering platform.

MOSS. An open-source tool-augmented conversational language model from Fudan University

stanford_alpaca. Code and documentation to train Stanford's Alpaca models, and generate the data.

self-instruct. Aligning pretrained language models with instruction data generated by themselves.

FastChat. An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.

LoRA. Code for loralib, an implementation of "LoRA: Low-Rank Adaptation of Large Language Models"

evals. Evals is a framework for evaluating LLMs and LLM systems, and an open-source registry of benchmarks.

openai-cookbook. Examples and guides for using the OpenAI API

IntroComputationModel. 南京大学宋方敏《计算模型导引》题解

NJU-TOC-Solutions. Solutions to "Introduction to Models of Computation" of Nanjing University

llama. Inference code for Llama models

fairscale. PyTorch extensions for high performance and large scale training.

DeepSpeed. DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

automl. Google Brain AutoML

attention-analysis. Jupyter Notebook

structural-probes. Codebase for testing whether hidden states of neural networks encode discrete structures.

Open-Assistant. OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.

StructAdapt. Structural Adapters in Pretrained Language Models for AMR-to-Text Generation (EMNLP 2021)

ATP. Source code for paper "ATP: AMRize Than Parse! Enhancing AMR Parsing with PseudoAMRs" @NAACL-2022

AMR-IE. The code repository for AMR guided joint information extraction model (NAACL-2021).

transition-amr-parser. SoTA Abstract Meaning Representation (AMR) parsing with word-node alignments in Pytorch. Includes checkpoints and other tools such as statistical significance Smatch.

amrlib. A python library that makes AMR parsing, generation and visualization simple.

BLINK. Entity Linker solution

cdec. Decoder, aligner, and model optimizer for statistical machine translation and other structured prediction models based on (mostly) context-free formalisms

AMR-Process. Code for preprocessing AMR graphs.

AMRBART. Code for our paper "Graph Pre-training for AMR Parsing and Generation" in ACL2022

penman. PENMAN notation (e.g. AMR) in Python

NJUNMT-pytorch. Python

perin. PERIN is Permutation-Invariant Semantic Parser developed for MRP 2020

camr. The system of SUDA-HUAWEI submitted at CAMR2022.

Two-Stage-CAMRP. Source code for paper "A Two-Stage Method for Chinese AMR Parsing" @ CAMRP-2022 & CCL-2022

spring. SPRING is a seq2seq model for Text-to-AMR and AMR-to-Text (AAAI2021).

kit. 🧱 Describe your site, AI builds it, you own it as Markdown. Snap together Tailwind blocks like Lego — landing pages, blogs, portfolios, docs & more. No AI slop. Free to deploy anywhere 👇

text2digits. Converts text such as "twenty three" to number/digit "23" in any sentence

pycnnum. Converting Chinese number string <=> int/float/str

w2n. Convert number words (eg. twenty one) to numeric digits (21)

cybertron-ai. mindspore implementation of transformers

fenci. 中文分词模块：继承了jieba分词的基本算法逻辑，进行了全方位的代码优化，还额外提供了HMM算法的训练功能支持。

PTR. Prompt Tuning with Rules

transformers. 🤗 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.

nju-health-report. 南京大学APP每日健康打卡。自用

English-to-IPA. Converts English text to IPA notation

nlp-datasets. Alphabetical list of free/public domain datasets with text data for use in Natural Language Processing (NLP)

DANN. pytorch implementation of Domain-Adversarial Training of Neural Networks

MyDeepLearning. A deep learning library to provide algs in pure Numpy or Tensorflow.

vaderSentiment. VADER Sentiment Analysis. VADER (Valence Aware Dictionary and sEntiment Reasoner) is a lexicon and rule-based sentiment analysis tool that is specifically attuned to sentiments expressed in social media, and works well on texts from other domains.