Beyond Output Matching: Bidirectional Alignment for Enhanced In-Context Learning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Qin, Chengwei, Xia, Wenhan, Jiao, Fangkai, Chen, Chen, Hu, Yuchen, Ding, Bosheng, Chen, Ruirui, Joty, Shafiq
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912406686400512
author Qin, Chengwei
Xia, Wenhan
Jiao, Fangkai
Chen, Chen
Hu, Yuchen
Ding, Bosheng
Chen, Ruirui
Joty, Shafiq
author_facet Qin, Chengwei
Xia, Wenhan
Jiao, Fangkai
Chen, Chen
Hu, Yuchen
Ding, Bosheng
Chen, Ruirui
Joty, Shafiq
contents Large language models (LLMs) have shown impressive few-shot generalization on many tasks via in-context learning (ICL). Despite their success in showing such emergent abilities, the scale and complexity of larger models also lead to unprecedentedly high computational demands and deployment challenges. In reaction, researchers explore transferring the powerful capabilities of larger models to more efficient and compact models by typically aligning the output of smaller (student) models with that of larger (teacher) models. Existing methods either train student models on the generated outputs of teacher models or imitate their token-level probability distributions. However, these distillation methods pay little to no attention to the input, which also plays a crucial role in ICL. Based on the finding that the performance of ICL is highly sensitive to the selection of demonstration examples, we propose Bidirectional Alignment (BiAlign) to fully leverage the models' preferences for ICL examples to improve the ICL abilities of student models. Specifically, we introduce the alignment of input preferences between student and teacher models by incorporating a novel ranking loss, in addition to aligning the token-level output distribution. With extensive experiments and analysis, we demonstrate that BiAlign can consistently outperform existing baselines on a variety of tasks involving language understanding, reasoning, and coding.
format Preprint
id arxiv_https___arxiv_org_abs_2312_17055
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Beyond Output Matching: Bidirectional Alignment for Enhanced In-Context Learning
Qin, Chengwei
Xia, Wenhan
Jiao, Fangkai
Chen, Chen
Hu, Yuchen
Ding, Bosheng
Chen, Ruirui
Joty, Shafiq
Computation and Language
Large language models (LLMs) have shown impressive few-shot generalization on many tasks via in-context learning (ICL). Despite their success in showing such emergent abilities, the scale and complexity of larger models also lead to unprecedentedly high computational demands and deployment challenges. In reaction, researchers explore transferring the powerful capabilities of larger models to more efficient and compact models by typically aligning the output of smaller (student) models with that of larger (teacher) models. Existing methods either train student models on the generated outputs of teacher models or imitate their token-level probability distributions. However, these distillation methods pay little to no attention to the input, which also plays a crucial role in ICL. Based on the finding that the performance of ICL is highly sensitive to the selection of demonstration examples, we propose Bidirectional Alignment (BiAlign) to fully leverage the models' preferences for ICL examples to improve the ICL abilities of student models. Specifically, we introduce the alignment of input preferences between student and teacher models by incorporating a novel ranking loss, in addition to aligning the token-level output distribution. With extensive experiments and analysis, we demonstrate that BiAlign can consistently outperform existing baselines on a variety of tasks involving language understanding, reasoning, and coding.
title Beyond Output Matching: Bidirectional Alignment for Enhanced In-Context Learning
topic Computation and Language
url https://arxiv.org/abs/2312.17055