• Publications
  • Influence
Language Models as Knowledge Bases?
TLDR
We present an in-depth analysis of the relational knowledge already present (without fine-tuning) in a wide range of state-of-the-art pretrained language models. Expand
  • 172
  • 32
  • PDF
MLQA: Evaluating Cross-lingual Extractive Question Answering
TLDR
We present MLQA, a multi-way aligned extractive QA evaluation benchmark intended to spur research in this area, and also provide machine-translation-based baselines. Expand
  • 60
  • 25
  • PDF
Dense Passage Retrieval for Open-Domain Question Answering
TLDR
We show that retrieval can be practically implemented using dense representations alone, where embeddings are learned from a small number of questions and passages by a simple dual-encoder framework. Expand
  • 55
  • 17
  • PDF
Interpretation of Natural Language Rules in Conversational Machine Reading
TLDR
We formalise this task and develop a crowd-sourcing strategy to collect 37k task instances based on real-world rules and crowd-generated questions and scenarios. Expand
  • 48
  • 11
  • PDF
Unsupervised Question Answering by Cloze Translation
TLDR
We propose and compare various unsupervised ways to perform cloze-to-natural question translation, including training an NMT model using non-aligned corpora of natural questions and cloze questions. Expand
  • 35
  • 8
  • PDF
How Context Affects Language Models' Factual Predictions
TLDR
We combine an information retrieval system with a pre-trained language model in a purely unsupervised way and show that the resulting system is competitive with a supervised machine reading baseline. Expand
  • 10
  • 2
  • PDF
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
TLDR
We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. Expand
  • 26
  • 1
  • PDF
Unsupervised Question Decomposition for Question Answering
TLDR
We propose an algorithm for One-to-N Unsupervised Sequence transduction (ONUS) that learns to map from the distribution of hard questions to that of many simple questions. Expand
  • 11
  • PDF
KILT: a Benchmark for Knowledge Intensive Language Tasks
Challenging problems such as open-domain question answering, fact checking, slot filling and entity linking require access to large, external knowledge sources. While some models do well onExpand
  • 5
  • PDF
Answering Complex Open-Domain Questions with Multi-Hop Dense Retrieval
TLDR
We propose a simple and efficient multi-hop dense retrieval approach for answering complex open-domain questions, which achieves state-of-the-art performance on two multi hop datasets, HotpotQA and multi-evidence FEVER. Expand
  • 2
  • PDF