The Superalignment of Superhuman Intelligence with Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Huang, Minlie, Wang, Yingkang, Cui, Shiyao, Ke, Pei, Tang, Jie
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912165042061312
author Huang, Minlie
Wang, Yingkang
Cui, Shiyao
Ke, Pei
Tang, Jie
author_facet Huang, Minlie
Wang, Yingkang
Cui, Shiyao
Ke, Pei
Tang, Jie
contents We have witnessed superhuman intelligence thanks to the fast development of large language models and multimodal language models. As the application of such superhuman models becomes more and more popular, a critical question arises here: how can we ensure superhuman models are still safe, reliable and aligned well to human values? In this position paper, we discuss the concept of superalignment from the learning perspective to answer this question by outlining the learning paradigm shift from large-scale pretraining, supervised fine-tuning, to alignment training. We define superalignment as designing effective and efficient alignment algorithms to learn from noisy-labeled data (point-wise samples or pair-wise preference data) in a scalable way when the task becomes very complex for human experts to annotate and the model is stronger than human experts. We highlight some key research problems in superalignment, namely, weak-to-strong generalization, scalable oversight, and evaluation. We then present a conceptual framework for superalignment, which consists of three modules: an attacker which generates adversary queries trying to expose the weaknesses of a learner model; a learner which will refine itself by learning from scalable feedbacks generated by a critic model along with minimal human experts; and a critic which generates critics or explanations for a given query-response pair, with a target of improving the learner by criticizing. We discuss some important research problems in each component of this framework and highlight some interesting research ideas that are closely related to our proposed framework, for instance, self-alignment, self-play, self-refinement, and more. Last, we highlight some future research directions for superalignment, including identification of new emergent risks and multi-dimensional alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2412_11145
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Superalignment of Superhuman Intelligence with Large Language Models
Huang, Minlie
Wang, Yingkang
Cui, Shiyao
Ke, Pei
Tang, Jie
Computation and Language
68T50
I.2.7
We have witnessed superhuman intelligence thanks to the fast development of large language models and multimodal language models. As the application of such superhuman models becomes more and more popular, a critical question arises here: how can we ensure superhuman models are still safe, reliable and aligned well to human values? In this position paper, we discuss the concept of superalignment from the learning perspective to answer this question by outlining the learning paradigm shift from large-scale pretraining, supervised fine-tuning, to alignment training. We define superalignment as designing effective and efficient alignment algorithms to learn from noisy-labeled data (point-wise samples or pair-wise preference data) in a scalable way when the task becomes very complex for human experts to annotate and the model is stronger than human experts. We highlight some key research problems in superalignment, namely, weak-to-strong generalization, scalable oversight, and evaluation. We then present a conceptual framework for superalignment, which consists of three modules: an attacker which generates adversary queries trying to expose the weaknesses of a learner model; a learner which will refine itself by learning from scalable feedbacks generated by a critic model along with minimal human experts; and a critic which generates critics or explanations for a given query-response pair, with a target of improving the learner by criticizing. We discuss some important research problems in each component of this framework and highlight some interesting research ideas that are closely related to our proposed framework, for instance, self-alignment, self-play, self-refinement, and more. Last, we highlight some future research directions for superalignment, including identification of new emergent risks and multi-dimensional alignment.
title The Superalignment of Superhuman Intelligence with Large Language Models
topic Computation and Language
68T50
I.2.7
url https://arxiv.org/abs/2412.11145