Reasons to Reject? Aligning Language Models with Judgments

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Weiwen, Cai, Deng, Zhang, Zhisong, Lam, Wai, Shi, Shuming
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910473619767296
author Xu, Weiwen
Cai, Deng
Zhang, Zhisong
Lam, Wai
Shi, Shuming
author_facet Xu, Weiwen
Cai, Deng
Zhang, Zhisong
Lam, Wai
Shi, Shuming
contents As humans, we consistently interact with our peers and receive feedback in the form of natural language. This language feedback allows us to maintain appropriate behavior, and rectify potential errors. The question arises naturally: can we use language feedback to align large language models (LLMs)? In contrast to previous research that aligns LLMs with scalar rewards, we present the first systematic exploration of alignment through the lens of language feedback (i.e., judgment). We start with an in-depth investigation of potential methods that can be adapted for aligning LLMs with judgments, revealing that these methods cannot fully capitalize on judgments. To facilitate more effective utilization of judgments, we propose a novel framework, Contrastive Unlikelihood Training (CUT), that allows for fine-grained inappropriate content detection and correction based on judgments. Our results show that, with merely 1317 off-the-shelf judgment data, CUT (LLaMA2-13b) can beat the 175B DaVinci003 and surpass the best baseline by 50.84 points on AlpacaEval. CUT (LLaMA2-chat-13b) can also align LLMs in an iterative fashion using up-to-date model-specific judgments, improving performance from 81.09 to 91.68 points on AlpacaEval. Further analysis suggests that judgments hold greater potential than rewards in LLM alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2312_14591
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Reasons to Reject? Aligning Language Models with Judgments
Xu, Weiwen
Cai, Deng
Zhang, Zhisong
Lam, Wai
Shi, Shuming
Computation and Language
As humans, we consistently interact with our peers and receive feedback in the form of natural language. This language feedback allows us to maintain appropriate behavior, and rectify potential errors. The question arises naturally: can we use language feedback to align large language models (LLMs)? In contrast to previous research that aligns LLMs with scalar rewards, we present the first systematic exploration of alignment through the lens of language feedback (i.e., judgment). We start with an in-depth investigation of potential methods that can be adapted for aligning LLMs with judgments, revealing that these methods cannot fully capitalize on judgments. To facilitate more effective utilization of judgments, we propose a novel framework, Contrastive Unlikelihood Training (CUT), that allows for fine-grained inappropriate content detection and correction based on judgments. Our results show that, with merely 1317 off-the-shelf judgment data, CUT (LLaMA2-13b) can beat the 175B DaVinci003 and surpass the best baseline by 50.84 points on AlpacaEval. CUT (LLaMA2-chat-13b) can also align LLMs in an iterative fashion using up-to-date model-specific judgments, improving performance from 81.09 to 91.68 points on AlpacaEval. Further analysis suggests that judgments hold greater potential than rewards in LLM alignment.
title Reasons to Reject? Aligning Language Models with Judgments
topic Computation and Language
url https://arxiv.org/abs/2312.14591