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

By Daniel Khashabi, Yeganeh Kordi, Hannaneh Hajishirzi. 2022

76 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Large Language Models, Question Answering & Reading Comprehension

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Large Language Models

In Question Answering & Reading Comprehension

Citations in the corpus

Plug-and-Play VQA: Zero-shot VQA by Conjoining Large Pretrained Models with Zero Training (2022, methodology): “Finally, we employ a PLM (Khashabi et al., 2022) to answer”
Rainier: Reinforced Knowledge Introspector for Commonsense Question Answering (2022, methodology): “For both imitation learning and reinforcement learning, we use the multiplechoice datasets that UnifiedQAv2 (Khashabi et al., 2022) uses for training (i.”
CONDAQA: A Contrastive Reading Comprehension Dataset for Reasoning about Negation (2022, methodology): “UniﬁedQA (Khashabi et al., 2020, 2022) Uni-ﬁedQA is built on top of the T5 architecture (Raffel et al., 2020) GPT-3 ( davinci ; Brown et al., 2020) This is the original GPT-3 model trained using only the standard LM objective.”
Rethinking with Retrieval: Faithful Large Language Model Inference (2022, methodology): “An alternative is to use an off-the-shelf model for inference with supporting facts, such as UniﬁedQA (Khashabi et al., 2020, 2022).”