We introduce ScaleQuest, a scalable, novel and cost-effective data synthesis method to unleash the reasoning capability of LLMs.

语言学竞赛集成 / Collection on Linguistics Olympiad (Chinese version only)

Must-read Papers on Knowledge Editing for Large Language Models.

A curated list of reinforcement learning with human feedback resources (continually updated)

An Easy-to-use, Scalable and High-performance RLHF Framework (70B+ PPO Full Tuning & Iterative DPO & LoRA & RingAttention & RFT)

Reference BLEU implementation that auto-downloads test sets and reports a version string to facilitate cross-lab comparisons

👨‍💻 An awesome and curated list of best code-LLM for research.

Google Research

Reference implementation for DPO (Direct Preference Optimization)

Code for "Learning to summarize from human feedback"

ChatGLM3 series: Open Bilingual Chat LLMs | 开源双语对话语言模型

Example models using DeepSpeed

A repo for distributed training of language models with Reinforcement Learning via Human Feedback (RLHF)

Fine-tuning ChatGLM-6B with PEFT | 基于 PEFT 的高效 ChatGLM 微调

Perspective is an API that uses machine learning models to score the perceived impact a comment might have on a conversation. See https://developers.perspectiveapi.com for more information.

Secrets of RLHF in Large Language Models Part I: PPO

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

OpenChat: Advancing Open-source Language Models with Imperfect Data

Fast and memory-efficient exact attention

ChatGLM2-6B: An Open Bilingual Chat LLM | 开源双语对话语言模型

基于ChatGLM-6B、ChatGLM2-6B、ChatGLM3-6B模型，进行下游具体任务微调，涉及Freeze、Lora、P-tuning、全参微调等

ChatGLM-6B: An Open Bilingual Dialogue Language Model | 开源双语对话语言模型

GLM-130B: An Open Bilingual Pre-Trained Model (ICLR 2023)

A trend starts from "Chain of Thought Prompting Elicits Reasoning in Large Language Models".

Code, datasets, and checkpoints for the paper "Improving Passage Retrieval with Zero-Shot Question Generation (EMNLP 2022)"

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

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

An original implementation of "MetaICL Learning to Learn In Context" by Sewon Min, Mike Lewis, Luke Zettlemoyer and Hannaneh Hajishirzi