LLMs：大型语言模型进化树结构图之模型(BERT-style/GPT-style)、数据(预训练数据/微调数据/测试数据)、NLP任务(五大任务+效率+可信度+基准指令调优+对齐)、三大类模型的使用和限制(Encoder-only、Encoder-Decoder、Decoder-only)

大型语言模型进化树结构图之模型(BERT-style/GPT-style)、数据(预训练数据/微调数据/测试数据)、NLP任务(五大任务+效率+可信度+基准指令调优+对齐)

大型语言模型进化树结构图

一、模型的实用指南：BERT-style/GPT-style

1.1、BERT-style 的语言模型：编码器-解码器或仅编码器

1.2、GPT-style的语言模型：仅解码器

二、数据实用指南：预训练数据/微调数据/测试数据

2.1、预训练数据

2.2、微调数据

2.3、测试数据/用户数据

三、自然语言处理任务的实用指南：NLP五大任务+效率+可信度+基准指令调优+对齐

3.1、传统的NLU任务

3.2、生成任务

3.3、知识密集型任务

3.4、随着规模增长的能力

3.5、特定任务

3.6、现实世界的任务

3.7、效率Efficiency

3.7.1、成本Cost

3.7.2、延迟Latency

3.7.3、参数高效微调Parameter-Efficient Fine-Tuning

3.7.4、预训练系统Pretraining System

3.8、可信度Trustworthiness

3.8.1、鲁棒性和校准性Robustness and Calibration

3.8.2、虚假偏见Spurious biases

3.8.3、安全问题Safety issues

3.9、基准指令调优Benchmark Instruction Tuning

3.10、对齐Alignment

3.10.1、安全对齐（无害）Safety Alignment (Harmless)

3.10.2、真实对齐（诚实） Truthfulness Alignment (Honest)

3.10.3、指导实践（有帮助）Practical Guides for Prompting (Helpful)

3.10.4、开源社区的对齐努力Alignment Efforts of Open-source Communtity

四、三大类模型的使用和限制：Encoder-only、Encoder-Decoder、Decoder-only

4.1、Encoder-only

4.2、Encoder-Decoder

4.3、Decoder-only

大型语言模型进化树结构图之模型(BERT-style/GPT-style)、数据(预训练数据/微调数据/测试数据)、NLP任务(五大任务+效率+可信度+基准指令调优+对齐)

大型语言模型进化树结构图

该列表基于我们的调研论文《在实践中利用LLMs的威力：关于ChatGPT及其后续发展的调研》，以及@xinyadu的努力。该调研部分基于这篇博客的后半部分。我们还构建了一个现代大型语言模型（LLMs）的演化树，以追踪近年来语言模型的发展，并突出了一些最知名的模型。
这些资源旨在帮助从业者在大型语言模型（LLMs）及其在自然语言处理（NLP）应用中的应用方面进行导航。我们还根据模型和数据许可信息包括它们的使用限制。如果您发现我们仓库中的任何资源有帮助，请随意使用（不要忘记引用我们的论文！��）。我们欢迎拉取请求来完善这个图表！

GitHub：GitHub - Mooler0410/LLMsPracticalGuide: A curated list of practical guide resources of LLMs (LLMs Tree, Examples, Papers)

一、模型的实用指南：BERT-style/GPT-style

1.1、BERT-style 的语言模型：编码器-解码器或仅编码器

BERT BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, 2018, Paper
RoBERTa RoBERTa: A Robustly Optimized BERT Pretraining Approach, 2019, Paper
DistilBERT DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter, 2019, Paper
ALBERT ALBERT: A Lite BERT for Self-supervised Learning of Language Representations, 2019, Paper
UniLM Unified Language Model Pre-training for Natural Language Understanding and Generation, 2019 Paper
ELECTRA ELECTRA: PRE-TRAINING TEXT ENCODERS AS DISCRIMINATORS RATHER THAN GENERATORS, 2020, Paper
T5 "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer". Colin Raffel et al. JMLR 2019. Paper
GLM "GLM-130B: An Open Bilingual Pre-trained Model". 2022. Paper
AlexaTM "AlexaTM 20B: Few-Shot Learning Using a Large-Scale Multilingual Seq2Seq Model". Saleh Soltan et al. arXiv 2022. Paper
ST-MoE ST-MoE: Designing Stable and Transferable Sparse Expert Models. 2022 Paper

BERT BERT：用于语言理解的深度双向转换器的预训练，2018年

RoBERTa RoBERTa：一种优化稳健的BERT预训练方法，2019年

DistilBERT DistilBERT：BERT的精简版本：更小、更快、更便宜、更轻巧，2019年

ALBERT ALBERT：一种轻量级的自监督学习语言表示方法，2019年

UniLM 统一语言模型预训练用于自然语言理解和生成，2019年

ELECTRA ELECTRA：将文本编码器作为判别器而非生成器进行预训练，2020年

T5 "探索具有统一文本到文本Transformer的迁移学习的极限"。Colin Raffel等人，JMLR 2019年

GLM "GLM-130B：一种开放的双语预训练模型"，2022年

AlexaTM "AlexaTM 20B：使用大规模多语言Seq2Seq模型进行少样本学习"。Saleh Soltan等人，arXiv 2022年

ST-MoE ST-MoE：设计稳定且可转移的稀疏专家模型。2022年

1.2、GPT-style的语言模型：仅解码器

GPT Improving Language Understanding by Generative Pre-Training. 2018. Paper
GPT-2 Language Models are Unsupervised Multitask Learners. 2018. Paper
GPT-3 "Language Models are Few-Shot Learners". NeurIPS 2020. Paper
OPT "OPT: Open Pre-trained Transformer Language Models". 2022. Paper
PaLM "PaLM: Scaling Language Modeling with Pathways". Aakanksha Chowdhery et al. arXiv 2022. Paper
BLOOM "BLOOM: A 176B-Parameter Open-Access Multilingual Language Model". 2022. Paper
MT-NLG "Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model". 2021. Paper
GLaM "GLaM: Efficient Scaling of Language Models with Mixture-of-Experts". ICML 2022. Paper
Gopher "Scaling Language Models: Methods, Analysis & Insights from Training Gopher". 2021. Paper
chinchilla "Training Compute-Optimal Large Language Models". 2022. Paper
LaMDA "LaMDA: Language Models for Dialog Applications". 2021. Paper
LLaMA "LLaMA: Open and Efficient Foundation Language Models". 2023. Paper
GPT-4 "GPT-4 Technical Report". 2023. Paper
BloombergGPT BloombergGPT: A Large Language Model for Finance, 2023, Paper
GPT-NeoX-20B: "GPT-NeoX-20B: An Open-Source Autoregressive Language Model". 2022. Paper

GPT 通过生成式预训练改进语言理解。2018年

GPT-2 语言模型是无监督多任务学习者。2018年

GPT-3 "语言模型是少样本学习者"。NeurIPS 2020年

OPT "OPT：开放预训练转换器语言模型"。2022年

PaLM "PaLM：通过路径扩展语言建模"。Aakanksha Chowdhery等人，arXiv 2022年

BLOOM "BLOOM：一种拥有176B参数的开放式多语言语言模型"。2022年

MT-NLG "使用DeepSpeed和Megatron训练Megatron-Turing NLG 530B，一种大规模生成语言模型"。2021年

GLaM "GLaM：通过专家混合实现语言模型的高效扩展"。ICML 2022年

Gopher "扩展语言模型：方法、分析和Gopher训练见解"。2021年

chinchilla "训练计算优化的大型语言模型"。2022年

LaMDA "LaMDA：用于对话应用的语言模型"。2021年

LLaMA "LLaMA：开放和高效的基础语言模型"。2023年

GPT-4 "GPT-4技术报告"。2023年

BloombergGPT BloombergGPT：一种面向金融领域的大型语言模型，2023年

GPT-NeoX-20B："GPT-NeoX-20B：一种开源自回归语言模型"。2022年

二、数据实用指南：预训练数据/微调数据/测试数据

2.1、预训练数据

RedPajama, 2023. Repo
The Pile: An 800GB Dataset of Diverse Text for Language Modeling, Arxiv 2020. Paper
How does the pre-training objective affect what large language models learn about linguistic properties?, ACL 2022. Paper
Scaling laws for neural language models, 2020. Paper
Data-centric artificial intelligence: A survey, 2023. Paper
How does GPT Obtain its Ability? Tracing Emergent Abilities of Language Models to their Sources, 2022. Blog

RedPajama，2023年，代码库

The Pile：一份用于语言建模的800GB多样化文本数据集，Arxiv 2020年

预训练目标如何影响大型语言模型对语言属性的学习？ACL 2022年

神经语言模型的扩展定律，2020年

以数据为中心的人工智能：一份调查，2023年

GPT如何获得其能力？追踪语言模型的新兴能力源自何处，2022年，博客

2.2、微调数据

Benchmarking zero-shot text classification: Datasets, evaluation and entailment approach, EMNLP 2019. Paper
Language Models are Few-Shot Learners, NIPS 2020. Paper
Does Synthetic Data Generation of LLMs Help Clinical Text Mining? Arxiv 2023 Paper

基于零样本文本分类的基准测试：数据集、评估和蕴含方法，EMNLP 2019年

语言模型是少样本学习者，NIPS 2020年

合成数据生成对临床文本挖掘有帮助吗？Arxiv 2023年

2.3、测试数据/用户数据

Shortcut learning of large language models in natural language understanding: A survey, Arxiv 2023. Paper
On the Robustness of ChatGPT: An Adversarial and Out-of-distribution Perspective Arxiv, 2023. Paper
SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems Arxiv 2019. Paper

自然语言理解中大型语言模型的快捷学习：一份调查，Arxiv 2023年

关于ChatGPT的鲁棒性：从对抗和分布外的角度看，Arxiv 2023年

SuperGLUE：一项用于通用语言理解系统的更具挑战性的基准测试，Arxiv 2019年

三、自然语言处理任务的实用指南：NLP五大任务+效率+可信度+基准指令调优+对齐

我们为用户的自然语言处理（NLP）应用程序构建了一个决策流程，用于选择LLMs或经过微调的模型\protect\footnotemark。该决策流程帮助用户评估他们手头的下游NLP应用程序是否满足特定条件，并基于该评估确定LLMs或经过微调的模型是否是他们应用程序的最佳选择。

3.1、传统的NLU任务

A benchmark for toxic comment classification on civil comments dataset Arxiv 2023 Paper
Is chatgpt a general-purpose natural language processing task solver? Arxiv 2023Paper
Benchmarking large language models for news summarization Arxiv 2022 Paper

Civil Comments数据集上有害评论分类的基准 Arxiv 2023

ChatGPT是否是一个通用的自然语言处理任务求解器？Arxiv 2023

大型语言模型在新闻摘要中的基准测试Arxiv 2022

3.2、生成任务

News summarization and evaluation in the era of gpt-3 Arxiv 2022 Paper
Is chatgpt a good translator? yes with gpt-4 as the engine Arxiv 2023 Paper
Multilingual machine translation systems from Microsoft for WMT21 shared task, WMT2021 Paper
Can ChatGPT understand too? a comparative study on chatgpt and fine-tuned bert, Arxiv 2023, Paper

在GPT-3时代的新闻摘要和评估 Arxiv 2022

ChatGPT是否是一个好的翻译器？是的，使用GPT-4作为引擎Arxiv 2023

微软的多语言机器翻译系统用于WMT21共享任务，WMT2021

ChatGPT能够理解吗？关于ChatGPT和经过微调的BERT的对比研究Arxiv 2023

3.3、知识密集型任务

Measuring massive multitask language understanding, ICLR 2021 Paper
Beyond the imitation game: Quantifying and extrapolating the capabilities of language models, Arxiv 2022 Paper
Inverse scaling prize, 2022 Link
Atlas: Few-shot Learning with Retrieval Augmented Language Models, Arxiv 2022 Paper
Large Language Models Encode Clinical Knowledge, Arxiv 2022 Paper

测量大规模多任务语言理解 ICLR 2021

超越模仿游戏：量化和推断语言模型的能力 Arxiv 2022

Inverse scaling prize, 2022 链接

Atlas: 带有检索增强的少样本学习语言模型Arxiv 2022

大型语言模型编码临床知识，Arxiv 2022

3.4、随着规模增长的能力

Training Compute-Optimal Large Language Models, NeurIPS 2022 Paper
Scaling Laws for Neural Language Models, Arxiv 2020 Paper
Solving math word problems with process- and outcome-based feedback, Arxiv 2022 Paper
Chain of thought prompting elicits reasoning in large language models, NeurIPS 2022 Paper
Emergent abilities of large language models, TMLR 2022 Paper
Inverse scaling can become U-shaped, Arxiv 2022 Paper
Towards Reasoning in Large Language Models: A Survey, Arxiv 2022 Paper

训练计算优化的大型语言模型 NeurIPS 2022

神经网络语言模型的缩放定律，Arxiv 2020

使用基于过程和结果的反馈解决数学问题，Arxiv 2022

思维链引导引发大型语言模型的推理，NeurIPS 2022

大型语言模型的新兴能力，TMLR 2022

Inverse scaling可能呈U形，Arxiv 2022

走向大型语言模型的推理：一项调查，Arxiv 2022

3.5、特定任务

Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks, Arixv 2022 Paper
PaLI: A Jointly-Scaled Multilingual Language-Image Model, Arxiv 2022 Paper
AugGPT: Leveraging ChatGPT for Text Data Augmentation, Arxiv 2023 Paper
Is gpt-3 a good data annotator?, Arxiv 2022 Paper
Want To Reduce Labeling Cost? GPT-3 Can Help, EMNLP findings 2021 Paper
GPT3Mix: Leveraging Large-scale Language Models for Text Augmentation, EMNLP findings 2021 Paper
LLM for Patient-Trial Matching: Privacy-Aware Data Augmentation Towards Better Performance and Generalizability, Arxiv 2023 Paper
ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks, Arxiv 2023 Paper
G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment, Arxiv 2023 Paper
GPTScore: Evaluate as You Desire, Arxiv 2023 Paper
Large Language Models Are State-of-the-Art Evaluators of Translation Quality, Arxiv 2023 Paper
Is ChatGPT a Good NLG Evaluator? A Preliminary Study, Arxiv 2023 Paper

图像作为外语：针对所有视觉和视觉-语言任务的BEiT预训练模型 Arixv 2022

PaLI：一个联合缩放的多语言语言-图像模型，Arxiv 2022

AugGPT：利用ChatGPT进行文本数据增强 Arxiv 2023

GPT-3是否是一个好的数据标注器？Arxiv 2022

想要降低标注成本吗？GPT-3可以帮助，EMNLP findings 2021

GPT3Mix：利用大规模语言模型进行文本增强，EMNLP findings 2021

LLM用于患者-试验匹配：隐私感知的数据增强以提高性能和普适性，Arxiv 2023

ChatGPT在文本注释任务中胜过众包工作者，Arxiv 2023

G-Eval：使用GPT-4进行自然语言生成评估，Arxiv 2023

GPTScore：根据需求进行评估，Arxiv 2023

大型语言模型是翻译质量的最先进评估器，Arxiv 2023

ChatGPT是一个好的自然语言生成评估器吗？初步研究，Arxiv 2023

3.6、现实世界的任务

Sparks of Artificial General Intelligence: Early experiments with GPT-4, Arxiv 2023 Paper

人工通用智能的火花：GPT-4的初步实验 Arxiv 2023

3.7、效率Efficiency

3.7.1、成本Cost

Openai’s gpt-3 language model: A technical overview, 2020. Blog Post
Measuring the carbon intensity of ai in cloud instances, FaccT 2022. Paper
In AI, is bigger always better?, Nature Article 2023. Article
Language Models are Few-Shot Learners, NeurIPS 2020. Paper
Pricing, OpenAI. Blog Post

OpenAI的GPT-3语言模型：技术概述，2020年博客文章

测量云计算实例中人工智能的碳强度，FaccT 2022

在人工智能领域，更大是否总是更好？，2023年自然文章

语言模型是少样本学习器，NeurIPS 2020

定价，OpenAI博客文章

3.7.2、延迟Latency

HELM: Holistic evaluation of language models, Arxiv 2022. Paper

HELM：语言模型的全面评估，Arxiv 2022

3.7.3、参数高效微调Parameter-Efficient Fine-Tuning

LoRA: Low-Rank Adaptation of Large Language Models, Arxiv 2021. Paper
Prefix-Tuning: Optimizing Continuous Prompts for Generation, ACL 2021. Paper
P-Tuning: Prompt Tuning Can Be Comparable to Fine-tuning Across Scales and Tasks, ACL 2022. Paper
P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Tasks, Arxiv 2022. Paper

LoRA：大型语言模型的低秩适应，Arxiv 2021

Prefix-Tuning：优化生成任务的连续提示，ACL 2021

P-Tuning：在规模和任务上，提示调优可以与微调相媲美，ACL 2022

P-Tuning v2：在各种规模和任务上，提示调优可以与微调普遍相媲美，Arxiv 2022

3.7.4、预训练系统Pretraining System

ZeRO: Memory Optimizations Toward Training Trillion Parameter Models, Arxiv 2019. Paper
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism, Arxiv 2019. Paper
Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM, Arxiv 2021. Paper
Reducing Activation Recomputation in Large Transformer Models, Arxiv 2021. Paper

ZeRO：向训练万亿参数模型进行内存优化，Arxiv 2019

Megatron-LM：使用模型并行性训练数十亿参数的语言模型，Arxiv 2019

使用Megatron-LM在GPU集群上进行高效的大规模语言模型训练，Arxiv 2021

减少大型Transformer模型中的激活重新计算，Arxiv 2021

3.8、可信度Trustworthiness

3.8.1、鲁棒性和校准性Robustness and Calibration

Calibrate before use: Improving few-shot performance of language models, ICML 2021. Paper
SPeC: A Soft Prompt-Based Calibration on Mitigating Performance Variability in Clinical Notes Summarization, Arxiv 2023. Paper

在使用之前进行校准：改善语言模型的少样本性能，ICML 2021

SPeC：软提示为基础的临床记录摘要中的校准，缓解性能变异性，Arxiv 2023

3.8.2、虚假偏见Spurious biases

Shortcut learning of large language models in natural language understanding: A survey, 2023 Paper
Mitigating gender bias in captioning system, WWW 2020 Paper
Calibrate Before Use: Improving Few-Shot Performance of Language Models, ICML 2021 Paper
Shortcut Learning in Deep Neural Networks, Nature Machine Intelligence 2020 Paper
Do Prompt-Based Models Really Understand the Meaning of Their Prompts?, NAACL 2022 Paper

自然语言理解中大型语言模型的快捷学习：一项调查，2023

减轻字幕系统中的性别偏见，WWW 2020

在使用之前进行校准：改善语言模型的少样本性能，ICML 2021

深度神经网络的快捷学习，Nature Machine Intelligence 2020

基于提示的模型真的理解其提示的含义吗？，NAACL 2022

3.8.3、安全问题Safety issues

GPT-4 System Card, 2023 Paper
The science of detecting llm-generated texts, Arxiv 2023 Paper
How stereotypes are shared through language: a review and introduction of the aocial categories and stereotypes communication (scsc) framework, Review of Communication Research, 2019 Paper
Gender shades: Intersectional accuracy disparities in commercial gender classification, FaccT 2018 Paper

GPT-4系统卡片，2023

检测LLM生成文本的科学，Arxiv 2023

通过语言分享刻板印象：社会类别和刻板印象传播（SCSC）框架的回顾和介绍，Review of Communication Research，2019

性别影子：商业性别分类中的交叉准确性差异，FaccT 2018

3.9、基准指令调优Benchmark Instruction Tuning

FLAN: Finetuned Language Models Are Zero-Shot Learners, Arxiv 2021 Paper
T0: Multitask Prompted Training Enables Zero-Shot Task Generalization, Arxiv 2021 Paper
Cross-task generalization via natural language crowdsourcing instructions, ACL 2022 Paper
Tk-INSTRUCT: Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks, EMNLP 2022 Paper
FLAN-T5/PaLM: Scaling Instruction-Finetuned Language Models, Arxiv 2022 Paper
The Flan Collection: Designing Data and Methods for Effective Instruction Tuning, Arxiv 2023 Paper
OPT-IML: Scaling Language Model Instruction Meta Learning through the Lens of Generalization, Arxiv 2023 Paper

FLAN：微调语言模型是零样本学习器，Arxiv 2021

T0：多任务提示训练实现零样本任务泛化，Arxiv 2021

通过自然语言众包指令实现跨任务泛化，ACL 2022

Tk-INSTRUCT：超自然指令：通过1600+ NLP任务的声明性指令实现泛化，EMNLP 2022

FLAN-T5/PaLM：扩展指令微调语言模型，Arxiv 2022

Flan集合：设计数据和方法以实现有效的指令调优，Arxiv 2023

OPT-IML：通过泛化视角扩展语言模型指令元学习，Arxiv 2023

3.10、对齐Alignment

Deep Reinforcement Learning from Human Preferences, NIPS 2017 Paper
Learning to summarize from human feedback, Arxiv 2020 Paper
A General Language Assistant as a Laboratory for Alignment, Arxiv 2021 Paper
Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback, Arxiv 2022 Paper
Teaching language models to support answers with verified quotes, Arxiv 2022 Paper
InstructGPT: Training language models to follow instructions with human feedback, Arxiv 2022 Paper
Improving alignment of dialogue agents via targeted human judgements, Arxiv 2022 Paper
Scaling Laws for Reward Model Overoptimization, Arxiv 2022 Paper
Scalable Oversight: Measuring Progress on Scalable Oversight for Large Language Models, Arxiv 2022 Paper

从人类偏好中进行深度强化学习，NIPS 2017

从人类反馈中学习总结，Arxiv 2020

作为对齐实验室的通用语言助手，Arxiv 2021

通过从人类反馈中进行强化学习训练一个有帮助且无害的助手，Arxiv 2022

教导语言模型通过验证的引用支持答案，Arxiv 2022

InstructGPT：通过人类反馈训练语言模型遵循指令，Arxiv 2022

通过有针对性的人类判断提高对话代理的对齐，Arxiv 2022

奖励模型过度优化的规模定律，Arxiv 2022

可扩展监督：衡量大型语言模型可扩展监督的进展，Arxiv 2022

3.10.1、安全对齐（无害）Safety Alignment (Harmless)

Red Teaming Language Models with Language Models, Arxiv 2022 Paper
Constitutional ai: Harmlessness from ai feedback, Arxiv 2022 Paper
The Capacity for Moral Self-Correction in Large Language Models, Arxiv 2023 Paper
OpenAI: Our approach to AI safety, 2023 Blog

使用语言模型对抗语言模型，Arxiv 2022

宪法AI：通过AI反馈实现无害性，Arxiv 2022

大型语言模型的道德自我修正能力，Arxiv 2023

OpenAI：我们对AI安全的方法，2023博客

3.10.2、真实对齐（诚实） Truthfulness Alignment (Honest)

Reinforcement Learning for Language Models, 2023 Blog

语言模型的强化学习，2023博客

3.10.3、指导实践（有帮助）Practical Guides for Prompting (Helpful)

OpenAI Cookbook. Blog
Prompt Engineering. Blog
ChatGPT Prompt Engineering for Developers! Course

OpenAI食谱。博客

提示工程。博客

开发人员的ChatGPT提示工程！课程

3.10.4、开源社区的对齐努力Alignment Efforts of Open-source Communtity

Self-Instruct: Aligning Language Model with Self Generated Instructions, Arxiv 2022 Paper
Alpaca. Repo
Vicuna. Repo
Dolly. Blog
DeepSpeed-Chat. Blog
GPT4All. Repo
OpenAssitant. Repo
ChatGLM. Repo
MOSS. Repo
Lamini. Repo/Blog

自我指导：用自动生成的指令对齐语言模型，Arxiv 2022

四、三大类模型的使用和限制：Encoder-only、Encoder-Decoder、Decoder-only

我们从模型及其预训练数据的角度提供信息。我们敦促社区中的用户参考公共模型和数据的许可信息，并负责任地使用它们。我们敦促开发者特别注意许可证的问题，使其透明且全面，以防止任何不必要和意外的使用情况。

4.1、Encoder-only

LLMs	Model			Data
	License	Commercial Use	Other noteable restrictions	License	Corpus
*BERT series of models (general domain)*	Apache 2.0	✅		Public	BooksCorpus, English Wikipedia
*RoBERTa*	MIT license	✅		Public	BookCorpus, CC-News, OpenWebText, STORIES
*ERNIE*	Apache 2.0	✅		Public	English Wikipedia
*SciBERT*	Apache 2.0	✅		Public	BERT corpus, 1.14M papers from Semantic Scholar
*LegalBERT*	CC BY-SA 4.0	❌		Public (except data from the Case Law Access Project)	EU legislation, US court cases, etc.
*BioBERT*	Apache 2.0	✅		PubMed	PubMed, PMC

4.2、Encoder-Decoder

	License	Commercial Use	Other noteable restrictions	License	Corpus
T5	Apache 2.0	✅		Public	C4
*Flan-T5*	Apache 2.0	✅		Public	C4, Mixture of tasks (Fig 2 in paper)
*BART*	Apache 2.0	✅		Public	RoBERTa corpus
*GLM*	Apache 2.0	✅		Public	BooksCorpus and English Wikipedia
*ChatGLM*	ChatGLM License	❌	No use for illegal purposes or military research, no harm the public interest of society	N/A	1T tokens of Chinese and English corpus

4.3、Decoder-only

	License	Commercial Use	Other noteable restrictions	License	Corpus
*GPT2*	Modified MIT License	✅	Use GPT-2 responsibly and clearly indicate your content was created using GPT-2.	Public	WebText
*GPT-Neo*	MIT license	✅		Public	Pile
*GPT-J*	Apache 2.0	✅		Public	Pile
*---> Dolly*	CC BY NC 4.0	❌		CC BY NC 4.0, Subject to terms of Use of the data generated by OpenAI	Pile, Self-Instruct
*---> GPT4ALL-J*	Apache 2.0	✅		Public	GPT4All-J dataset
*Pythia*	Apache 2.0	✅		Public	Pile
*---> Dolly v2*	MIT license	✅		Public	Pile, databricks-dolly-15k
*OPT*	OPT-175B LICENSE AGREEMENT	❌	No development relating to surveillance research and military, no harm the public interest of society	Public	RoBERTa corpus, the Pile, PushShift.io Reddit
*---> OPT-IML*	OPT-175B LICENSE AGREEMENT	❌	same to OPT	Public	OPT corpus, Extended version of Super-NaturalInstructions
*YaLM*	Apache 2.0	✅		Unspecified	Pile, Teams collected Texts in Russian
*BLOOM*	The BigScience RAIL License	✅	No use of generating verifiably false information with the purpose of harming others; content without expressly disclaiming that the text is machine generated	Public	ROOTS corpus (Lauren¸con et al., 2022)
*---> BLOOMZ*	The BigScience RAIL License	✅	same to BLOOM	Public	ROOTS corpus, xP3
*Galactica*	CC BY-NC 4.0	❌		N/A	The Galactica Corpus
*LLaMA*	Non-commercial bespoke license	❌	No development relating to surveillance research and military, no harm the public interest of society	Public	CommonCrawl, C4, Github, Wikipedia, etc.
*---> Alpaca*	CC BY NC 4.0	❌		CC BY NC 4.0, Subject to terms of Use of the data generated by OpenAI	LLaMA corpus, Self-Instruct
*---> Vicuna*	CC BY NC 4.0	❌		Subject to terms of Use of the data generated by OpenAI; Privacy Practices of ShareGPT	LLaMA corpus, 70K conversations from ShareGPT.com
*---> GPT4ALL*	GPL Licensed LLaMa	❌		Public	GPT4All dataset
*OpenLLaMA*	Apache 2.0	✅		Public	RedPajama
*CodeGeeX*	The CodeGeeX License	❌	No use for illegal purposes or military research	Public	Pile, CodeParrot, etc.
*StarCoder*	BigCode OpenRAIL-M v1 license	✅	No use of generating verifiably false information with the purpose of harming others; content without expressly disclaiming that the text is machine generated	Public	The Stack
*MPT-7B*	Apache 2.0	✅		Public	mC4 (english), The Stack, RedPajama, S2ORC
*falcon*	TII Falcon LLM License	✅/❌	Available under a license allowing commercial use	Public	RefinedWeb