Détail du document
Identifiant
oai:arXiv.org:2406.01306
Sujet
Computer Science - Computation and...Auteur
Qu, Fanyi Sun, Hao Wu, YunfangCatégorie
Computer Science
Année
2024
Date de référencement
05/06/2024
Mots clés
language model distractor dgRésumé
Within the context of reading comprehension, the task of Distractor Generation (DG) aims to generate several incorrect options to confuse readers.
Traditional supervised methods for DG rely heavily on expensive human-annotated distractor labels.
In this paper, we propose an unsupervised DG framework, leveraging Large Language Models (LLMs) as cost-effective annotators to enhance the DG capability of smaller student models.
Specially, to perform knowledge distilling, we propose a dual task training strategy that integrates pseudo distractors from LLMs and the original answer in-formation as the objective targets with a two-stage training process.
Moreover, we devise a counterfactual contrastive decoding mechanism for increasing the distracting capability of the DG model.
Experiments show that our unsupervised generation method with Bart-base greatly surpasses GPT-3.5-turbo performance with only 200 times fewer model parameters.
Our proposed unsupervised DG method offers a cost-effective framework for practical reading comprehension applications, without the need of laborious distractor annotation and costly large-size models ;Comment: Accepted as a long paper in ACL 2024 findings
Qu, Fanyi,Sun, Hao,Wu, Yunfang, 2024, Unsupervised Distractor Generation via Large Language Model Distilling and Counterfactual Contrastive Decoding
High-Frequency Repetitive Magnetic Stimulation at the Sacrum Alleviates Chronic Constipation in Parkinson’s Patients
magnetic stimulation parkinson’s significant patients scale sacrum pd hf-rms chronic constipation scores
