UnifiedQA-v2: Stronger Generalization via Broader Cross-Format Training

We present UnifiedQA-v2, a QA model built with the same process as UnifiedQA, except that it utilizes more supervision -- roughly 3x the number of datasets used for UnifiedQA. This generally leads to better in-domain and cross-domain results.

cs.CL cs.AI Topic Modeling Natural Language Processing Techniques Machine Learning and Data Classification

Source

Paper

Similar papers

UNIFIEDQA: Crossing Format Boundaries with a Single QA SystemDaniel Khashabi, Sewon Min, Tushar Khot et al.2020 In Library

Learning to Generalize for Cross-domain QAYingjie Niu, Linyi Yang, Ruihai Dong et al.2023 · arXiv (Cornell University) · 8 citations In Library

Learning to Generalize for Cross-domain QAYingjie Niu, Linyi Yang, Ruihai Dong et al.2023 In Library

Multi-Task Learning and Domain-Specific Models to Improve Robustness of QA SystemAnmol Gupta, Meredith Xu2021 · 0 citations In Library

Generalizable Video Quality Assessment via Weak-to-Strong LearningLinhan Cao, Wei Sun, Xiangyang Zhu et al.2025 · 7 citations In Library

ProQA: Structural Prompt-based Pre-training for Unified Question AnsweringWanjun Zhong, Yifan Gao, Ning Ding et al.2022 · arXiv (Cornell University) · 39 citations In Library

CrossVQA: Scalably Generating Benchmarks for Systematically Testing VQA GeneralizationArjun Akula, Soravit Changpinyo, Boqing Gong et al.2021 · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · 30 citations In Library

M2QA: Multi-domain Multilingual Question AnsweringLeon Engländer, Hannah Sterz, Clifton Poth et al.2024 · arXiv (Cornell University) · 6 citations In Library

Unified Question Generation with Continual Lifelong LearningWei Yuan, Hongzhi Yin, Tieke He et al.2022 · The Web Conference · 13 citations In Library

M2QA: Multi-domain Multilingual Question AnsweringLeon Engländer, Hannah Sterz, Clifton Poth et al.2024 In Library

Patents near this paper

Unsupervised Real-to-Virtual Domain Unification for End-to-End Highway DrivingUS 11,543,830 · PETUUM, INC. · 2023 EXPANDING TRAINING QUESTIONS THROUGH CONTEXTUALIZING FEATURE SEARCHUS 11,017,312 · INTERNATIONAL BUSINESS MACHINES CORPORATION · 2021 METHOD FOR QA WITH MULTI-MODAL INFORMATIONUS 11,720,750 · ACTIONPOWER CORP. · 2023 DETECTING DUPLICATED QUESTIONS USING REVERSE GRADIENT ADVERSARIAL DOMAIN ADAPTATIONUS 10,983,971 · INTUIT INC. · 2021 LEARNING APPROXIMATE TRANSLATIONS OF UNFAMILIAR MEASUREMENT UNITS DURING DEEP QUESTION ANSWERING SYSTEM TRAINING AND USAGEUS 11,556,758 · INTERNATIONAL BUSINESS MACHINES CORPORATION · 2023

Source of record

arXiv