We explore supervised multitask pre-training by proposing Instruction Pre-Training, a framework that scalably augments massive raw corpora with instruction-response pairs to pre-train language models. The instruction-response pairs are generated by an efficient instruction synthesizer built on open-source models. In our experiments, we synthesize 200M instruction-response pairs covering 40+ task categories to verify the effectiveness of Instruction Pre-Training. Instruction Pre-Training* outperforms Vanilla Pre-training in both general pre-training from scratch and domain-adaptive continual pre-training. In pre-training from scratch, Instruction Pre-Training not only improves pre-trained base models but also benefits more from further instruction tuning. In continual pre-training, Instruction Pre-Training enables Llama3-8B to be comparable to or even outperform Llama3-70B.

**************************** Updates ****************************

2026/1/23: Released LLM-in-Sandbox Elicits General Agentic Intelligence, where the data of Instruction Pre-Training achieves robust generalization in agentic RL!
2024/11/30: Released the multimodal version of the instruction synthesizer: Visual Instruction Synthesizer
2024/9/20: Our paper has been accepted by EMNLP 2024 main conference🎉
2024/9/11: Updated FAQ on continual pre-training from Llama3
2024/8/29: Updated guidelines on evaluating any 🤗Huggingface models on the domain-specific tasks
2024/7/31: Updated pre-training suggestions in the Advanced Usage section of instruction-synthesizer
2024/7/15: We scaled up the pre-trained tokens from 100B to 250B, with the number of synthesized instruction-response pairs reaching 500M. The performance trend on downstream tasks throughout the pre-training process:

* 2024/6/21: Released the [paper](https://huggingface.co/papers/2406.14491), [code](https://github.com/microsoft/LMOps), and [resources](https://huggingface.co/instruction-pretrain)

Resources

🤗 We share our data and models with example usages, feel free to open any discussions at this page! 🤗

Thanks to the demo davanstrien/instruction-synthesizer for implementing our approach
Context-Based Instruction Synthesizer: instruction-synthesizer
Fine-Tuning Data for the Synthesizer: ft-instruction-synthesizer-collection
General Models Pre-Trained from Scratch (on 100B tokes):
- InstructLM-500M
- InstructLM-1.3B
Domain-Specific Models Pre-Trained from Llama3-8B:
- Finance-Llama3-8B
- Biomedicine-Llama3-8B
General Instruction-Augmented Corpora: general-instruction-augmented-corpora
Domain-Specific Instruction-Augmented Corpora (no finance data to avoid ethical issues): medicine-instruction-augmented-corpora

General Pre-Training From Scratch

We augment the RefinedWeb corproa with instruction-response pairs generated by our context-based instruction synthesizer to pre-train general langauge models from scratch.

To evaluate our general base model using the lm-evaluation-harness framework

Setup dependencies:

bash

git clone https://github.com/EleutherAI/lm-evaluation-harness
cd lm-evaluation-harness
pip install -e .

Evaluate:

bash

MODEL=instruction-pretrain/InstructLM-1.3B
add_bos_token=True # this flag is needed because lm-eval-harness set add_bos_token to False by default, but ours require add_bos_token to be True

accelerate launch -m lm_eval --model hf \
    --model_args pretrained=${MODEL},add_bos_token=${add_bos_token},dtype=float16  \
    --gen_kwargs do_sample=False \
    --tasks piqa,hellaswag,winogrande \
    --batch_size auto \
    --num_fewshot 0

accelerate launch -m lm_eval --model hf \
    --model_args pretrained=${MODEL},add_bos_token=${add_bos_token},dtype=float16 \
    --gen_kwargs do_sample=False \
    --tasks social_iqa,ai2_arc,openbookqa,boolq,mmlu \
    --batch_size auto \
    --num_fewshot 5

Citation

If you find our work helpful, please cite us:

Instruction Pre-Training (EMNLP 2024)

bibtex

@article{cheng2024instruction,
  title={Instruction Pre-Training: Language Models are Supervised Multitask Learners},
  author={Cheng, Daixuan and Gu, Yuxian and Huang, Shaohan and Bi, Junyu and Huang, Minlie and Wei, Furu},
  journal={arXiv preprint arXiv:2406.14491},
  year={2024}
}

Adapt LLM to Domains（ICLR 2024）

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}
}