Pretraining Language Models with Human Preferences

Tomasz Korbak, Kejian Shi, Angelica Chen, Rasika Bhalerao, Christopher L. Buckley, Jason Phang, Samuel R. Bowman, Ethan Perez

    Research output: Contribution to journalConference articlepeer-review


    Language models (LMs) are pretrained to imitate internet text, including content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, and more. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the trade-off between alignment and capabilities of pretrained LMs. We find a Pareto-optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores given by a reward model. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversariallychosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training.

    Original languageEnglish (US)
    Pages (from-to)17506-17533
    Number of pages28
    JournalProceedings of Machine Learning Research
    StatePublished - 2023
    Event40th International Conference on Machine Learning, ICML 2023 - Honolulu, United States
    Duration: Jul 23 2023Jul 29 2023

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Software
    • Control and Systems Engineering
    • Statistics and Probability


    Dive into the research topics of 'Pretraining Language Models with Human Preferences'. Together they form a unique fingerprint.

    Cite this