Headline

# Adapting Large Language Models to Domains via Continual Pre-Training
This repo contains the **Headline dataset** used in our **ICLR 2024** paper [Adapting Large Language Models via Reading Comprehension](https://huggingface.co/papers/2309.09530).

We explore **continued pre-training on domain-specific corpora** for large language models. While this approach enriches LLMs with domain knowledge, it significantly hurts their prompting ability for question answering. Inspired by human learning via reading comprehension, we propose a simple method to **transform large-scale pre-training corpora into reading comprehension texts**, consistently improving prompting performance across tasks in biomedicine, finance, and law domains. **Our 7B model competes with much larger domain-specific models like BloombergGPT-50B**.

### [2024/11/29] 🤗 Introduce the multimodal version of AdaptLLM at [AdaMLLM](https://huggingface.co/AdaptLLM/Adapt-MLLM-to-Domains), for adapting MLLMs to domains 🤗

**************************** **Updates** ****************************
* 2024/11/29: Released [AdaMLLM](https://huggingface.co/AdaptLLM/Adapt-MLLM-to-Domains) for adapting MLLMs to domains
* 2024/9/20: Our [research paper for Instruction-Pretrain](https://huggingface.co/papers/2406.14491) has been accepted by EMNLP 2024
* 2024/8/29: Updated [guidelines](https://huggingface.co/datasets/AdaptLLM/finance-tasks) on evaluating any 🤗Huggingface models on the domain-specific tasks
* 2024/6/22: Released the [benchmarking code](https://github.com/microsoft/LMOps/tree/main/adaptllm)
* 2024/6/21: Released the general version of AdaptLLM at [Instruction-Pretrain](https://huggingface.co/instruction-pretrain)
* 2024/4/2: Released the [raw data splits (train and test)](https://huggingface.co/datasets/AdaptLLM/ConvFinQA) of all the evaluation datasets
* 2024/1/16: Our [research paper for AdaptLLM](https://huggingface.co/papers/2309.09530) has been accepted by ICLR 2024
* 2023/12/19: Released our [13B base models](https://huggingface.co/AdaptLLM/law-LLM-13B) developed from LLaMA-1-13B
* 2023/12/8: Released our [chat models](https://huggingface.co/AdaptLLM/law-chat) developed from LLaMA-2-Chat-7B
* 2023/9/18: Released our [paper](https://huggingface.co/papers/2309.09530), [code](https://github.com/microsoft/LMOps), [data](https://huggingface.co/datasets/AdaptLLM/law-tasks), and [base models](https://huggingface.co/AdaptLLM/law-LLM) developed from LLaMA-1-7B

## Domain-Specific LLaMA-1
### LLaMA-1-7B
In our paper, we develop three domain-specific models from LLaMA-1-7B, which are also available in Huggingface: [Biomedicine-LLM](https://huggingface.co/AdaptLLM/medicine-LLM), [Finance-LLM](https://huggingface.co/AdaptLLM/finance-LLM) and [Law-LLM](https://huggingface.co/AdaptLLM/law-LLM), the performances of our AdaptLLM compared to other domain-specific LLMs are:

<p align='center'>
<img src="https://cdn-uploads.huggingface.co/production/uploads/650801ced5578ef7e20b33d4/6efPwitFgy-pLTzvccdcP.png" width="700">
</p>

### LLaMA-1-13B
Moreover, we scale up our base model to LLaMA-1-13B to see if **our method is similarly effective for larger-scale models**, and the results are consistently positive too: [Biomedicine-LLM-13B](https://huggingface.co/AdaptLLM/medicine-LLM-13B), [Finance-LLM-13B](https://huggingface.co/AdaptLLM/finance-LLM-13B) and [Law-LLM-13B](https://huggingface.co/AdaptLLM/law-LLM-13B).

## Domain-Specific LLaMA-2-Chat
Our method is also effective for aligned models! LLaMA-2-Chat requires a [specific data format](https://huggingface.co/blog/llama2#how-to-prompt-llama-2), and our **reading comprehension can perfectly fit the data format** by transforming the reading comprehension into a multi-turn conversation. We have also open-sourced chat models in different domains: [Biomedicine-Chat](https://huggingface.co/AdaptLLM/medicine-chat), [Finance-Chat](https://huggingface.co/AdaptLLM/finance-chat) and [Law-Chat](https://huggingface.co/AdaptLLM/law-chat)

## Domain-Specific Tasks

### Pre-templatized/Formatted Testing Splits
To easily reproduce our prompting results, we have uploaded the filled-in zero/few-shot input instructions and output completions of the test each domain-specific task: [biomedicine-tasks](https://huggingface.co/datasets/AdaptLLM/medicine-tasks), [finance-tasks](https://huggingface.co/datasets/AdaptLLM/finance-tasks), and [law-tasks](https://huggingface.co/datasets/AdaptLLM/law-tasks).

**Note:** those filled-in instructions are specifically tailored for models before alignment and do NOT fit for the specific data format required for chat models.

### Raw Datasets
We have also uploaded the raw training and testing splits, for facilitating fine-tuning or other usages:
- [ChemProt](https://huggingface.co/datasets/AdaptLLM/ChemProt)
- [RCT](https://huggingface.co/datasets/AdaptLLM/RCT)
- [ConvFinQA](https://huggingface.co/datasets/AdaptLLM/ConvFinQA)
- [FiQA_SA](https://huggingface.co/datasets/AdaptLLM/FiQA_SA)
- [Headline](https://huggingface.co/datasets/AdaptLLM/Headline)
- [NER](https://huggingface.co/datasets/AdaptLLM/NER)
- [FPB](https://huggingface.co/datasets/AdaptLLM/FPB)

The other datasets used in our paper have already been available in huggingface, and you can directly load them with the following code:
```python
from datasets import load_dataset

# MQP:
dataset = load_dataset('medical_questions_pairs')
# PubmedQA:
dataset = load_dataset('bigbio/pubmed_qa')
# USMLE:
dataset=load_dataset('GBaker/MedQA-USMLE-4-options')
# SCOTUS
dataset = load_dataset("lex_glue", 'scotus')
# CaseHOLD
dataset = load_dataset("lex_glue", 'case_hold')
# UNFAIR-ToS
dataset = load_dataset("lex_glue", 'unfair_tos')
```

## Citation
If you find our work helpful, please cite us:
```bibtex
@inproceedings{
cheng2024adapting,
title={Adapting Large Language Models via Reading Comprehension},
author={Daixuan Cheng and Shaohan Huang and Furu Wei},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=y886UXPEZ0}
}
```

and the original dataset:
```bibtex
@article{Headline,
author = {Ankur Sinha and
Tanmay Khandait},
title = {Impact of News on the Commodity Market: Dataset and Results},
journal = {CoRR},
volume = {abs/2009.04202},
year = {2020}
}
```

# 通过持续预训练使大语言模型（Large Language Model, LLM）适配特定领域
本仓库包含我们发表于**ICLR 2024**的论文《通过阅读理解适配大语言模型》（[Adapting Large Language Models via Reading Comprehension](https://huggingface.co/papers/2309.09530)）中使用的**Headline数据集**。

我们探索了针对大语言模型在领域专属语料上的持续预训练方案。尽管该方法可为LLM注入领域知识，但会显著损害其在问答任务中的提示学习能力。受人类通过阅读理解进行学习的启发，我们提出了一种简单方法，可将大规模预训练语料转换为阅读理解文本，该方法能持续提升生物医学、金融与法律领域各类任务的提示学习性能。我们的7B参数模型可与BloombergGPT-50B等超大规模领域专属模型相媲美。

### [2024/11/29] 🤗 在[AdaMLLM](https://huggingface.co/AdaptLLM/Adapt-MLLM-to-Domains)推出AdaptLLM的多模态版本，用于适配多模态大语言模型（Multimodal Large Language Model, MLLM）到特定领域 🤗

**************************** **更新日志** ****************************
* 2024/11/29：发布用于领域适配的[AdaMLLM](https://huggingface.co/AdaptLLM/Adapt-MLLM-to-Domains)
* 2024/9/20：我们的《指令预训练》研究论文（[Instruction-Pretrain](https://huggingface.co/papers/2406.14491)）已被EMNLP 2024收录
* 2024/8/29：更新了在领域专属任务上评估🤗Huggingface模型的[指南](https://huggingface.co/datasets/AdaptLLM/finance-tasks)
* 2024/6/22：发布基准测试代码（[benchmarking code](https://github.com/microsoft/LMOps/tree/main/adaptllm)）
* 2024/6/21：在[Instruction-Pretrain](https://huggingface.co/instruction-pretrain)发布AdaptLLM的通用版本
* 2024/4/2：发布所有评估数据集的原始数据划分（训练集与测试集）[ConvFinQA](https://huggingface.co/datasets/AdaptLLM/ConvFinQA)
* 2024/1/16：我们的AdaptLLM研究论文（[AdaptLLM](https://huggingface.co/papers/2309.09530)）已被ICLR 2024收录
* 2023/12/19：发布基于LLaMA-1-13B开发的[13B基础模型](https://huggingface.co/AdaptLLM/law-LLM-13B)
* 2023/12/8：发布基于LLaMA-2-Chat-7B开发的[对话模型](https://huggingface.co/AdaptLLM/law-chat)
* 2023/9/18：发布我们的论文（[paper](https://huggingface.co/papers/2309.09530)）、代码（[code](https://github.com/microsoft/LMOps)）、数据集（[data](https://huggingface.co/datasets/AdaptLLM/law-tasks)）以及基于LLaMA-1-7B开发的[基础模型](https://huggingface.co/AdaptLLM/law-LLM)

## 领域专属LLaMA-1
### LLaMA-1-7B
在我们的论文中，我们基于LLaMA-1-7B开发了三款领域专属模型，均已在Huggingface平台开源：[生物医学大模型（Biomedicine-LLM）](https://huggingface.co/AdaptLLM/medicine-LLM)、[金融大模型（Finance-LLM）](https://huggingface.co/AdaptLLM/finance-LLM)与[法律大模型（Law-LLM）](https://huggingface.co/AdaptLLM/law-LLM)。我们的AdaptLLM与其他领域专属大语言模型的性能对比如下：

<p align='center'>
<img src="https://cdn-uploads.huggingface.co/production/uploads/650801ced5578ef7e20b33d4/6efPwitFgy-pLTzvccdcP.png" width="700">
</p>

### LLaMA-1-13B
此外，我们将基础模型扩展至LLaMA-1-13B，以验证我们的方法在更大规模模型上是否同样有效，实验结果同样表现优异：[生物医学大模型-13B（Biomedicine-LLM-13B）](https://huggingface.co/AdaptLLM/medicine-LLM-13B)、[金融大模型-13B（Finance-LLM-13B）](https://huggingface.co/AdaptLLM/finance-LLM-13B)与[法律大模型-13B（Law-LLM-13B）](https://huggingface.co/AdaptLLM/law-LLM-13B)。

## 领域专属LLaMA-2-Chat
我们的方法对对齐后的模型同样有效！LLaMA-2-Chat需要特定的数据格式（[specific data format](https://huggingface.co/blog/llama2#how-to-prompt-llama-2)），而我们的阅读理解任务可通过将文本转换为多轮对话，完美适配该数据格式。我们还开源了不同领域的对话模型：[生物医学对话模型（Biomedicine-Chat）](https://huggingface.co/AdaptLLM/medicine-chat)、[金融对话模型（Finance-Chat）](https://huggingface.co/AdaptLLM/finance-chat)与[法律对话模型（Law-Chat）](https://huggingface.co/AdaptLLM/law-chat)

## 领域专属任务
### 预模板化/格式化测试划分
为便于复现我们的提示学习实验结果，我们上传了各领域专属任务测试集的填充式零样本/少样本输入指令与输出结果：[生物医学任务集（biomedicine-tasks）](https://huggingface.co/datasets/AdaptLLM/medicine-tasks)、[金融任务集（finance-tasks）](https://huggingface.co/datasets/AdaptLLM/finance-tasks)与[法律任务集（law-tasks）](https://huggingface.co/datasets/AdaptLLM/law-tasks)。

**注意：** 此类填充式指令专为未对齐的模型设计，不适用于对话模型所需的专属数据格式。

### 原始数据集
我们还上传了原始训练与测试划分，以支持微调或其他应用场景：
- [ChemProt](https://huggingface.co/datasets/AdaptLLM/ChemProt)
- [RCT](https://huggingface.co/datasets/AdaptLLM/RCT)
- [ConvFinQA](https://huggingface.co/datasets/AdaptLLM/ConvFinQA)
- [FiQA_SA](https://huggingface.co/datasets/AdaptLLM/FiQA_SA)
- [Headline](https://huggingface.co/datasets/AdaptLLM/Headline)
- [NER](https://huggingface.co/datasets/AdaptLLM/NER)
- [FPB](https://huggingface.co/datasets/AdaptLLM/FPB)

我们论文中使用的其他数据集已在Huggingface平台上线，可通过以下代码直接加载：
python
from datasets import load_dataset

# 医学问题对数据集（MQP）:
dataset = load_dataset('medical_questions_pairs')
# PubMed问答数据集（PubmedQA）:
dataset = load_dataset('bigbio/pubmed_qa')
# USMLE考试数据集:
dataset=load_dataset('GBaker/MedQA-USMLE-4-options')
# 美国最高法院案例数据集（SCOTUS）:
dataset = load_dataset("lex_glue", 'scotus')
# CaseHOLD数据集:
dataset = load_dataset("lex_glue", 'case_hold')
# 不公平服务条款数据集（UNFAIR-ToS）:
dataset = load_dataset("lex_glue", 'unfair_tos')

## 引用
若您的工作受益于本项目，请引用如下文献：
bibtex
@inproceedings{
cheng2024adapting,
title={Adapting Large Language Models via Reading Comprehension},
author={Daixuan Cheng and Shaohan Huang and Furu Wei},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=y886UXPEZ0}
}

以及原始数据集的引用：
bibtex
@article{Headline,
author = {Ankur Sinha and
Tanmay Khandait},
title = {Impact of News on the Commodity Market: Dataset and Results},
journal = {CoRR},
volume = {abs/2009.04202},
year = {2020}
}