Papperoni

2022

Training Language Models to Follow Instructions With Human Feedback

Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, C. Wainwright, Pamela Mishkin, Chiyuan Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F. Christiano, Jan Leike, Ryan Lowe

Open PDF Google Scholar

citations

Cite Score

AI summary

This paper introduces InstructGPT models, which are fine-tuned GPT-3 models using human feedback, showing improved alignment with user intent, better truthfulness, reduced toxicity, and superior performance on various NLP tasks compared to the original GPT-3, despite having significantly fewer parameters.

Main Contributions

Introduced InstructGPT models, fine-tuned GPT-3 models with human feedback for aligning language models with user intent.
Utilized supervised learning on labeler demonstrations and reinforcement learning from human feedback on ranked model outputs for fine-tuning.
Demonstrated that 1.3B parameter InstructGPT models are preferred over 175B GPT-3 outputs in human evaluations.
Showed improvements in truthfulness and reductions in toxic output generation with InstructGPT models.
Mitigated performance regressions on public NLP datasets by mixing pretraining gradients during RLHF fine-tuning (PPO-ptx models).

Abstract

Making language models bigger does not inherently make them better at following a user’s intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent.

Citation Graph

Loading graph...

References [91]

Sort:

Filter:

[1]Language Models Are Few-Shot Learners

Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, Dario Amodei - 2020

21 papers in library cite