Conditional Language Learning with Context

Xiao Zhang, Miao Li, Ji Wu
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:59247-59263, 2024.

Abstract

Language models can learn sophisticated language understanding skills from fitting raw text. They also unselectively learn useless corpus statistics and biases, especially during finetuning on domain-specific corpora. In this paper, we propose a simple modification to causal language modeling called conditional finetuning, which performs language modeling conditioned on a context. We show that a context can "explain away" certain corpus statistics and make the model avoid learning them. In this fashion, conditional finetuning achieves selective learning from a corpus, learning knowledge useful for downstream tasks while avoiding learning useless corpus statistics like topic biases. This selective learning effect leads to less forgetting and better stability-plasticity tradeoff in domain finetuning, potentially benefitting lifelong learning with language models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-zhang24ag, title = {Conditional Language Learning with Context}, author = {Zhang, Xiao and Li, Miao and Wu, Ji}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {59247--59263}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/zhang24ag/zhang24ag.pdf}, url = {https://proceedings.mlr.press/v235/zhang24ag.html}, abstract = {Language models can learn sophisticated language understanding skills from fitting raw text. They also unselectively learn useless corpus statistics and biases, especially during finetuning on domain-specific corpora. In this paper, we propose a simple modification to causal language modeling called conditional finetuning, which performs language modeling conditioned on a context. We show that a context can "explain away" certain corpus statistics and make the model avoid learning them. In this fashion, conditional finetuning achieves selective learning from a corpus, learning knowledge useful for downstream tasks while avoiding learning useless corpus statistics like topic biases. This selective learning effect leads to less forgetting and better stability-plasticity tradeoff in domain finetuning, potentially benefitting lifelong learning with language models.} }
Endnote
%0 Conference Paper %T Conditional Language Learning with Context %A Xiao Zhang %A Miao Li %A Ji Wu %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-zhang24ag %I PMLR %P 59247--59263 %U https://proceedings.mlr.press/v235/zhang24ag.html %V 235 %X Language models can learn sophisticated language understanding skills from fitting raw text. They also unselectively learn useless corpus statistics and biases, especially during finetuning on domain-specific corpora. In this paper, we propose a simple modification to causal language modeling called conditional finetuning, which performs language modeling conditioned on a context. We show that a context can "explain away" certain corpus statistics and make the model avoid learning them. In this fashion, conditional finetuning achieves selective learning from a corpus, learning knowledge useful for downstream tasks while avoiding learning useless corpus statistics like topic biases. This selective learning effect leads to less forgetting and better stability-plasticity tradeoff in domain finetuning, potentially benefitting lifelong learning with language models.
APA
Zhang, X., Li, M. & Wu, J.. (2024). Conditional Language Learning with Context. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:59247-59263 Available from https://proceedings.mlr.press/v235/zhang24ag.html.

Related Material