---
title: "London — research institutions and their papers"
description: "London (United Kingdom): 14 institutions with papers on Elamur Research — 8 universities, 1 research institute, 4 companies, 1 nonprofit — and 119,289…"
canonical: "https://research.elamur.ai/free/en/agglomerations/london/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/agglomerations/london/)

[home](https://md.research.elamur.ai/free/en/) / [institutions](https://md.research.elamur.ai/free/en/orgs/) / [agglomerations](https://md.research.elamur.ai/free/en/agglomerations/) / London

# London

Agglomeration [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/) · metro population 12,598,186 (functional urban area, GHSL 2015)

**14** institutions with a page**189** works in our library**119,289** citations of their works

## About

### Profile

London is a city region in the [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/). On this site it appears as a hub for language technology and machine learning. Its research comes from universities, an independent research institute, the London offices of large technology companies and a cultural nonprofit. Two universities sit outside the urban core but belong to the wider region: the [University of Surrey](https://md.research.elamur.ai/free/en/universities/university-of-surrey/) in Guildford and the [University of Hertfordshire](https://md.research.elamur.ai/free/en/universities/university-of-hertfordshire/) in Hatfield.

The most cited institutions here mix industry and academia. [Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/) comes first, followed by [University College London](https://md.research.elamur.ai/free/en/universities/university-college-london/), [Meta (United Kingdom)](https://md.research.elamur.ai/free/en/companies/meta-united-kingdom/) and [Imperial College London](https://md.research.elamur.ai/free/en/universities/imperial-college-london/). After them come [Queen Mary University of London](https://md.research.elamur.ai/free/en/universities/queen-mary-university-of-london/), [The Alan Turing Institute](https://md.research.elamur.ai/free/en/labs/the-alan-turing-institute/) and [King's College London](https://md.research.elamur.ai/free/en/universities/king-s-college-london/). The region also includes [SOAS University of London](https://md.research.elamur.ai/free/en/universities/soas-university-of-london/) and the [University of London](https://md.research.elamur.ai/free/en/universities/university-of-london/). Further industry presence comes from [Systems, Applications & Products in Data Processing](https://md.research.elamur.ai/free/en/companies/systems-applications-products-in-data-processing/) and [Google (United Kingdom)](https://md.research.elamur.ai/free/en/companies/google-united-kingdom/). The [Goethe-Institute United Kingdom](https://md.research.elamur.ai/free/en/ngos/goethe-institute-united-kingdom/) is the nonprofit among them.

Most of the region's work in our library is natural language processing. It covers language-understanding benchmarks, question answering, machine translation, fact checking, abusive language detection, coreference and anaphora, and the behaviour of large language models. A smaller group of works deals with deep generative modelling, code generation and protein structure prediction.

### History

Among the works here, the region's research begins with rule-based computational linguistics. The earliest is [An Algorithm for Pronominal Anaphora Resolution](https://md.research.elamur.ai/free/en/papers/lappin-1994-algorithm-pronominal-anaphora-resolution/) from 1994, associated with SOAS, which presents RAP, a procedure that resolves pronouns through hand-built rules. Reference stayed in view afterwards. A 2000 study connected to the University of London asked what MUC-style coreference annotation really captures, and a 2004 King's College London paper detected verb phrase ellipsis in automatically parsed text. In 2006 the University of London appears on a shared-task report comparing manual and automatic evaluation of translation between European languages.

The 2010s brought statistical and then neural methods. A 2012 Hertfordshire framework took up compositionality in distributional semantics. In 2014 come a study of deep latent Gaussian models under Google DeepMind, a definition of fact checking as a task connected to UCL, and a King's College London method for identifying the languages within multilingual documents. From 2015 machine reading, stance detection, sequence labelling, aspect-based sentiment analysis and open-source neural translation appear.

The late 2010s are the busiest stretch, with benchmarks for general language understanding, multilingual sentence embeddings, cross-lingual question answering, the FEVER claim-verification task and language models probed as stores of knowledge. In the 2020s attention turns to large language models: emergent abilities, prompt order, clinical knowledge, model collapse, role-play, mathematical reasoning and code generation. The newest works carry the year 2026; one adapts metaphor identification to Arabic, the other examines grounding in dialogue fact verification.

### Research profile

General-purpose language understanding and evaluation is the most visible strand. [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://md.research.elamur.ai/free/en/papers/wang-2018-glue-multi-task-benchmark-analysis/) gathered several tasks into one platform for measuring progress, and [SuperGLUE](https://md.research.elamur.ai/free/en/papers/wang-2019-superglue-stickier-benchmark-general-purpose/) followed with harder English problems; both appear under Google DeepMind. Evaluation continues later. [Humanity's Last Exam](https://md.research.elamur.ai/free/en/papers/phan-2025-humanity-s-last-exam/), from 2025, responds to the near-saturation of benchmarks such as MMLU and is shared by Imperial, Queen Mary, the Turing Institute and Hertfordshire. BigCodeBench, linked to Queen Mary, moves code evaluation beyond short standalone exercises.

Question answering and knowledge-intensive work centres on UCL and Meta (United Kingdom). [Teaching Machines to Read and Comprehend](https://md.research.elamur.ai/free/en/papers/hermann-2015-teaching-machines-read-comprehend/) answered the lack of large supervised reading corpora by pairing documents with questions and answers. [Language Models as Knowledge Bases?](https://md.research.elamur.ai/free/en/papers/petroni-2019-language-models-as-knowledge-bases/) asked whether pretrained models could stand in for costly structured knowledge bases. KILT gathered knowledge-intensive tasks under one benchmark with a library of retrievers and readers, and MLQA offered cross-lingual extractive question answering in seven languages. A UCL study showed that BERT models could link entities never seen in training. At Meta, [Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering](https://md.research.elamur.ai/free/en/papers/izacard-2021-leveraging-passage-retrieval-generative-models/) introduced Fusion-in-Decoder, whose decoder reads the joined encodings of many retrieved passages.

Multilingual processing and machine translation run across several institutions. At Meta (United Kingdom), Artetxe and Schwenk built one sentence encoder shared by many language families and scripts; WikiMatrix mined parallel Wikipedia sentences for pairs that often exclude English; CCNet turned raw Common Crawl snapshots into clean monolingual corpora; and LayerDrop, a dropout that skips whole layers, made Transformer depth adjustable. At Imperial, [OpenNMT](https://md.research.elamur.ai/free/en/papers/klein-2017-opennmt-open-source-toolkit-neural/) released an open neural translation framework. Surrey adds a line on quality estimation: [TransQuest](https://md.research.elamur.ai/free/en/papers/ranasinghe-2020-transquest-translation-quality-estimation-cross/) built it on cross-lingual Transformers, and Surrey appears on the WMT quality-estimation findings for 2022 and 2023, on the second WMT task on sign language translation, and on a study of cyber risks from critical translation errors in Arabic mental-health tweets.

Fact checking and misinformation connect UCL, Imperial, King's, Queen Mary and the Turing Institute. The 2014 UCL task definition framed fact checking as ordinal classification. The FEVER shared task, under Imperial, targeted claim verification against evidence, and FEVEROUS, under King's, extended it to structured sources. Imperial's explainable fact checking for public health built the PUBHEALTH claims collection. RumourEval, shared by Queen Mary and the Turing Institute, set a benchmark for rumour veracity and support. UCL's stance detection on tweets with bidirectional conditional encoding belongs here too, and a 2024 King's study carries stance detection into multi-modal posts.

Online harms and subjective annotation are a strength of the Turing Institute and Queen Mary. A 2019 survey set out open problems in abusive content detection; Learning from the Worst trained hate-speech classifiers on adversarially generated rounds; [HateCheck](https://md.research.elamur.ai/free/en/papers/rottger-2021-hatecheck-functional-tests-hate-speech/) borrowed functional testing from software engineering to diagnose such classifiers; and a 2022 paper asked dataset creators to choose between two annotation paradigms. Queen Mary's works include a position paper against relying on one aggregated gold label and a crowdsourced anaphora corpus that keeps competing readings. The University of London adds ArMIS, an Arabic misogyny and sexism corpus recording annotator disagreement, and Queen Mary hosts the SemEval task on explainable sexism detection.

Coreference and structured prediction form the oldest continuous line, running from the 1994 algorithm through the 2000 critique and the 2004 ellipsis detector to a neural model for bridging reference resolution and Queen Mary's named entity recognition as dependency parsing, which handles nested and flat entities in one system. Meta (United Kingdom) trained a character-level language model jointly with a sequence tagger. Systems, Applications & Products in Data Processing contributes aspect-based sentiment analysis through recursive neural conditional random fields and an interactive multi-task network.

Large language models are the newest strand. [Emergent Abilities of Large Language Models](https://md.research.elamur.ai/free/en/papers/wei-2022-emergent-abilities-large-language-models/) questioned the picture of smooth gains with scale. Large Language Models Encode Clinical Knowledge proposed instruction prompt tuning for medicine, and AlphaCode adapted Transformers trained on GitHub code to programming contests. UCL showed that the order of in-context examples matters greatly. Imperial's works include a conceptual paper on role-play and the curse-of-recursion study of training on generated data. Meta explained fine-tuning through intrinsic dimensionality, the Turing Institute's MetaMath bootstrapped mathematical questions, and Surrey's Selective Context removes redundant input to speed inference. Imperial's [PEGASUS](https://md.research.elamur.ai/free/en/papers/zhang-2019-pegasus-pre-training-extracted-gap/) pretrained summarisation models by generating gap sentences.

Beyond language, [Highly accurate protein structure prediction with AlphaFold](https://md.research.elamur.ai/free/en/papers/jumper-2021-highly-accurate-protein-structure-prediction/) brings in structural biology, and the 2014 stochastic backpropagation study represents probabilistic machine learning. King's College London adds medical text through a corpus annotating patients, interventions and outcomes and through paragraph-level simplification of medical writing, as well as Masader Plus for exploring Arabic datasets.

### Leading institutions and labs

Google DeepMind is the most cited institution in the region on this site, with works from GLUE and machine reading to deep generative models, clinical language models, AlphaCode and AlphaFold. University College London has the most works here, centred on question answering, knowledge probing, entity linking, explanation-based inference through e-SNLI, fact checking and stance detection. Meta (United Kingdom) concentrates on multilingual representation, web-scale corpora, retrieval-augmented question answering and efficient Transformers. Imperial College London spans translation tooling, summarisation, fact checking, sarcasm detection, model collapse and frontier evaluation.

Queen Mary University of London works on coreference, annotation disagreement and shared tasks on rumours and sexism. The Alan Turing Institute covers online harms, annotation design, lexical semantic change and mathematical reasoning. King's College London covers clinical text, structured fact verification, numerical reasoning and multilingual documents. SOAS joins the 1994 anaphora algorithm with the Arabic metaphor guideline. The University of London contributes translation evaluation, coreference theory, ArMIS, bridging resolution and dialogue fact verification. Surrey is the region's centre for quality estimation, Hertfordshire contributes distributional semantics, and Systems, Applications & Products in Data Processing contributes sentiment analysis.

### Leading scientists

At Google DeepMind the first names are [Olaf Ronneberger](https://md.research.elamur.ai/free/en/authors/olaf-ronneberger/), Anna Potapenko, Oriol Vinyals, Koray Kavukcuoglu, Felix Hill, Chris Dyer, Phil Blunsom and Sebastian Ruder, whose institution's work here runs from protein structure to language benchmarks and code generation.

At University College London the group includes Jacob Menick, [Sebastian Riedel](https://md.research.elamur.ai/free/en/authors/sebastian-riedel/), Patrick Lewis, Martin Chadwick, Tim Rocktäschel, Yuxiang Wu, Timothy Lillicrap and Pontus Stenetorp, in a setting of knowledge probing, KILT, MLQA and zero-shot entity linking.

At Meta (United Kingdom), [Holger Schwenk](https://md.research.elamur.ai/free/en/authors/holger-schwenk/) comes first; Schwenk's name is on the multilingual sentence encoder and on WikiMatrix. Edouard Grave, a co-author of LayerDrop, follows, then Vishrav Chaudhary, Francisco Guzman, Fabio Petroni, Guillaume Lample, Juan Pino and Sonal Gupta.

Imperial College London begins with [Lucia Specia](https://md.research.elamur.ai/free/en/authors/lucia-specia/), followed by Shamsuddeen Hassan Muhammad, Marek Rei, Oana Cocarascu, Pranava Madhyastha, Zhenhao Li, Alexander M. Rush and Yue Zhang, amid work on translation, summarisation and fact checking.

At Queen Mary, [Massimo Poesio](https://md.research.elamur.ai/free/en/authors/massimo-poesio/) comes first, with a name on the crowdsourced anaphora corpus that preserves disagreement. Juntao Yu follows, linked to entity recognition as dependency parsing and to bridging resolution, then Silviu Paun, Alexandra Uma, Arkaitz Zubiaga, Abdulrahman Aloraini, Dirk Hovy and Shalom Lappin.

The Alan Turing Institute's group opens with Charles Sutton and [Bertie Vidgen](https://md.research.elamur.ai/free/en/authors/bertie-vidgen/), whose work here covers abusive content detection and dynamically generated hate-speech data. Walid Magdy, Dong Nguyen and Thomas Lukasiewicz follow, then [Paul Rottger](https://md.research.elamur.ai/free/en/authors/paul-rottger/), behind HateCheck and the annotation-paradigms paper, with Scott A. Hale and Helen Margetts.

King's College London begins with [Helen Yannakoudakis](https://md.research.elamur.ai/free/en/authors/helen-yannakoudakis/), then Michael Schlichtkrull, Shalom Lappin, Iain Marshall, Oana Cocarascu, Yulan He, Lin Gui and Elena Simperl. [Shalom Lappin](https://md.research.elamur.ai/free/en/authors/shalom-lappin/) also heads the SOAS group with Rodger Kibble and Mandar Marathe; Lappin's 1994 algorithm is the region's earliest work here, and Marathe's name is on the 2026 metaphor guideline. Rodger Kibble also opens the University of London group, alongside Christof Monz, Mikel Artetxe, Jean Maillard, David Dale, Colin Cherry, Adam Roberts and Andy Jones.

At Surrey, [Constantin Orasan](https://md.research.elamur.ai/free/en/authors/constantin-orasan/) comes first, then Yucheng Li, Diptesh Kanojia, Pushpak Bhattacharyya, Chenghua Lin, David Corney and Hui Wang. At Systems, Applications & Products in Data Processing, [Daniel Dahlmeier](https://md.research.elamur.ai/free/en/authors/daniel-dahlmeier/) heads a group with Wee Sun Lee, Manan Dey, Hwee Tou Ng, Ruidan He and Wenya Wang. At Hertfordshire, [Daoud Clarke](https://md.research.elamur.ai/free/en/authors/daoud-clarke/) is linked to the 2012 compositionality framework, and Andy Jones also appears.

### Landmark works

GLUE is the most cited of the region's works among papers in our library; it turned scattered language-understanding tasks into a shared yardstick. Teaching Machines to Read and Comprehend follows, having made supervised reading comprehension possible through large question–answer corpora. Emergent Abilities of Large Language Models argued that some capabilities appear abruptly as models grow. SuperGLUE raised the bar with harder tasks, and Language Models as Knowledge Bases? showed that pretrained models hold relational knowledge recoverable with cloze-style queries.

Close behind come works that shaped tooling and method: PEGASUS for summarisation, OpenNMT as an open translation framework, Fusion-in-Decoder for retrieval-based answering, Fantastically Ordered Prompts on the sensitivity of in-context learning, the Meta multilingual sentence encoder, AlphaCode, MetaMath and the clinical-knowledge study. AlphaFold stands apart as the region's main contribution outside language, a redesigned deep network that predicts protein structure from sequences and their alignments.

### Collaborations and links between industry and academia

Several works here are shared across London institutions. MLQA joins UCL and Meta (United Kingdom). Humanity's Last Exam joins Imperial, Queen Mary, the Turing Institute and Hertfordshire. RumourEval is shared by Queen Mary and the Turing Institute. The BigScience ROOTS corpus, a multilingual collection built by an open workshop, links Queen Mary with Systems, Applications & Products in Data Processing.

People also cross institutional lines. Shalom Lappin appears at Queen Mary, King's and SOAS; Oana Cocarascu at Imperial and King's; Rodger Kibble at SOAS and the University of London; Andy Jones at the University of London and Hertfordshire. Mikel Artetxe, co-author of the Meta sentence encoder, also appears in the University of London group, and OpenNMT, under Imperial, came from authors at SYSTRAN and Harvard NLP. Broadly, the companies supply much of the most cited benchmark and model work here, while the universities and the Turing Institute carry fact checking, annotation, online harms and translation quality.

### Timeline

From 1994 to 2006 the works here are few and linguistic: rule-based pronoun resolution, the nature of coreference annotation, verb phrase ellipsis and evaluation of European-language translation.

Between 2012 and 2016 come distributional semantics, deep generative models, fact checking as a task, language identification in mixed documents, machine reading, stance detection and neural sentiment analysis.

From 2017 to 2019 output grows quickly, with OpenNMT, GLUE, SuperGLUE, e-SNLI, FEVER, multilingual sentence embeddings, WikiMatrix, MLQA, knowledge probing, PEGASUS, LayerDrop, RumourEval and the survey of abusive content detection.

From 2020 to 2022 come KILT, entity linking, CCNet, TransQuest, nested entity recognition, AlphaFold, Fusion-in-Decoder, HateCheck, FEVEROUS, emergent abilities, clinical knowledge, AlphaCode, prompt order and ArMIS.

From 2023 to 2026 the focus is large language models and their evaluation: model collapse, role-play, MetaMath, context compression, numerical reasoning, ROOTS, sexism detection, quality estimation, sign language translation, BigCodeBench and Humanity's Last Exam, closing with the Arabic metaphor guideline and grounding in dialogue fact verification.

## Institutions (14)

8 universities, 1 research institute, 4 companies, 1 nonprofit.

### 🎓 Universities (8)

1. **[University College London](https://md.research.elamur.ai/free/en/universities/university-college-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*44 works · 14,853 citations*
2. **[Imperial College London](https://md.research.elamur.ai/free/en/universities/imperial-college-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*21 works · 8,081 citations*
3. **[Queen Mary University of London](https://md.research.elamur.ai/free/en/universities/queen-mary-university-of-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*27 works · 2,331 citations*
4. **[King's College London](https://md.research.elamur.ai/free/en/universities/king-s-college-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*13 works · 894 citations*
5. **[SOAS University of London](https://md.research.elamur.ai/free/en/universities/soas-university-of-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*2 works · 945 citations*
6. **[University of London](https://md.research.elamur.ai/free/en/universities/university-of-london/)**University · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*5 works · 759 citations*
7. **[University of Surrey](https://md.research.elamur.ai/free/en/universities/university-of-surrey/)**University · Guildford, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*6 works · 561 citations*
8. **[University of Hertfordshire](https://md.research.elamur.ai/free/en/universities/university-of-hertfordshire/)**University · Hatfield, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*2 works · 318 citations*

### 🔬 Labs and research institutes (1)

1. **[The Alan Turing Institute](https://md.research.elamur.ai/free/en/labs/the-alan-turing-institute/)**Research institute · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*13 works · 2,969 citations*

### 🏢 Companies (4)

1. **[Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/)**Company · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*29 works · 74,823 citations*
2. **[Meta (United Kingdom)](https://md.research.elamur.ai/free/en/companies/meta-united-kingdom/)**Company · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*36 works · 11,073 citations*
3. **[Systems, Applications & Products in Data Processing](https://md.research.elamur.ai/free/en/companies/systems-applications-products-in-data-processing/)**Company · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*3 works · 1,007 citations*
4. **[Google (United Kingdom)](https://md.research.elamur.ai/free/en/companies/google-united-kingdom/)**Company · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*5 works · 674 citations*

### 🤝 Nonprofits (1)

1. **[Goethe-Institute United Kingdom](https://md.research.elamur.ai/free/en/ngos/goethe-institute-united-kingdom/)**Nonprofit · London, [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/)*3 works · 1 citation*

## Works in our library (189)

Works of its institutions, the most cited first.

1. [**Highly accurate protein structure prediction with AlphaFold**John Jumper, Richard Evans, Alexander Pritzel, et al.*2021 · 43,470 citations*](https://md.research.elamur.ai/free/en/papers/jumper-2021-highly-accurate-protein-structure-prediction/)
2. [**GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding**Alex Wang, Amanpreet Singh, Julian Michael, et al.*2018 · 7,141 citations*](https://md.research.elamur.ai/free/en/papers/wang-2018-glue-multi-task-benchmark-analysis/)
3. [**Large Language Models Encode Clinical Knowledge**Karan Singhal, Shekoofeh Azizi, Tao Tu, et al.*2022 · 4,464 citations*](https://md.research.elamur.ai/free/en/papers/singhal-2022-large-language-models-encode-clinical/)
4. [**Teaching Machines to Read and Comprehend**Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, et al.*2015 · 3,806 citations*](https://md.research.elamur.ai/free/en/papers/hermann-2015-teaching-machines-read-comprehend/)
5. [**Language Models as Knowledge Bases?**Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, et al.*2019 · 2,977 citations*](https://md.research.elamur.ai/free/en/papers/petroni-2019-language-models-as-knowledge-bases/)
6. [**Stochastic Backpropagation and Approximate Inference in Deep Generative Models**Danilo Jimenez Rezende, Shakir Mohamed, Daan Wierstra*2014 · 2,586 citations*](https://md.research.elamur.ai/free/en/papers/rezende-2014-stochastic-backpropagation-approximate-inference-deep/)
7. [**OpenNMT: Open-Source Toolkit for Neural Machine Translation**Guillaume Klein, Yoon Kim, Yuntian Deng, et al.*2017 · 2,460 citations*](https://md.research.elamur.ai/free/en/papers/klein-2017-opennmt-open-source-toolkit-neural/)
8. [**Emergent Abilities of Large Language Models**Jason Wei, Yi Tay, Rishi Bommasani, et al.*2022 · 2,215 citations*](https://md.research.elamur.ai/free/en/papers/wei-2022-emergent-abilities-large-language-models/)
9. [**SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems**Alex Wang, Yada Pruksachatkun, Nikita Nangia, et al.*2019 · 2,187 citations*](https://md.research.elamur.ai/free/en/papers/wang-2019-superglue-stickier-benchmark-general-purpose/)
10. [**PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization**Jingqing Zhang, Yao Zhao, Mohammad Saleh, et al.*2019 · 1,806 citations*](https://md.research.elamur.ai/free/en/papers/zhang-2019-pegasus-pre-training-extracted-gap/)
11. [**Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering**Gautier Izacard, Edouard Grave*2021 · 1,674 citations*](https://md.research.elamur.ai/free/en/papers/izacard-2021-leveraging-passage-retrieval-generative-models/)
12. [**Competition-Level Code Generation with AlphaCode**Yujia Li, David Choi, Junyoung Chung, et al.*2022 · 1,543 citations*](https://md.research.elamur.ai/free/en/papers/li-2022-competition-level-code-generation-alphacode/)
13. [**Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity**Yao Lu, Max Bartolo, Alastair Moore, et al.*2022 · 1,187 citations*](https://md.research.elamur.ai/free/en/papers/lu-2022-fantastically-ordered-prompts-where-find/)
14. [**The NarrativeQA Reading Comprehension Challenge**Tomáš Kočiský, Jonathan Schwarz, Phil Blunsom, et al.*2018 · 1,026 citations*](https://md.research.elamur.ai/free/en/papers/kocisky-2018-narrativeqa-reading-comprehension-challenge/)
15. [**Universal and Transferable Adversarial Attacks on Aligned Language Models**Andy Zou, Zifan Wang, Nicholas Carlini, et al.*2023 · 951 citations*](https://md.research.elamur.ai/free/en/papers/zou-2023-universal-transferable-adversarial-attacks-aligned/)
16. [**An Algorithm for Pronominal Anaphora Resolution**Shalom Lappin, Herbert J. Leass*1994 · 945 citations*](https://md.research.elamur.ai/free/en/papers/lappin-1994-algorithm-pronominal-anaphora-resolution/)
17. [**Topic Modeling in Embedding Spaces**Adji B. Dieng, Francisco J. R. Ruiz, David M. Blei*2020 · 914 citations*](https://md.research.elamur.ai/free/en/papers/dieng-2020-topic-modeling-embedding-spaces/)
18. [**MAD-X: An Adapter-Based Framework for Multi-Task Cross-Lingual Transfer**Jonas Pfeiffer, Ivan Vulić, Iryna Gurevych, et al.*2020 · 831 citations*](https://md.research.elamur.ai/free/en/papers/pfeiffer-2020-mad-x-adapter-based-framework/)
19. [**MLQA: Evaluating Cross-lingual Extractive Question Answering**Patrick Lewis, Barlas Oğuz, Ruty Rinott, et al.*2019 · 735 citations*](https://md.research.elamur.ai/free/en/papers/lewis-2019-mlqa-evaluating-cross-lingual-extractive/)
20. [**KILT: a Benchmark for Knowledge Intensive Language Tasks**Fabio Petroni, Aleksandra Piktus, Angela Fan, et al.*2020 · 663 citations*](https://md.research.elamur.ai/free/en/papers/petroni-2020-kilt-benchmark-knowledge-intensive-language/)
21. [**CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data**Guillaume Wenzek, Marie-Anne Lachaux, Alexis Conneau, et al.*2020 · 650 citations*](https://md.research.elamur.ai/free/en/papers/wenzek-2020-ccnet-extracting-high-quality-monolingual/)
22. [**Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyond**Mikel Artetxe, Holger Schwenk*2018 · 627 citations*](https://md.research.elamur.ai/free/en/papers/artetxe-2018-massively-multilingual-sentence-embeddings-zero/)
23. [**Scalable Zero-shot Entity Linking with Dense Entity Retrieval**Ledell Wu, Fabio Petroni, Martin Josifoski, et al.*2020 · 618 citations*](https://md.research.elamur.ai/free/en/papers/wu-2020-scalable-zero-shot-entity-linking/)
24. [**e-SNLI: Natural Language Inference with Natural Language Explanations**Oana-Maria Camburu, Tim Rocktäschel, Thomas Lukasiewicz, et al.*2018 · 612 citations*](https://md.research.elamur.ai/free/en/papers/camburu-2018-e-snli-natural-language-inference/)
25. [**Role-Play with Large Language Models**Murray Shanahan, Kyle McDonell, Laria Reynolds*2023 · 606 citations*](https://md.research.elamur.ai/free/en/papers/shanahan-2023-role-play-large-language-models/)

[See all 189 works →](https://md.research.elamur.ai/free/en/agglomerations/london/works/)

## Works by year

### 2020s

[**2026** 11 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2026/) · [**2025** 4 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2025/) · [**2024** 8 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2024/) · [**2023** 24 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2023/) · [**2022** 24 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2022/) · [**2021** 22 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2021/) · [**2020** 34 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2020/)

### 2010s

[**2019** 26 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2019/) · [**2018** 14 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2018/) · [**2017** 5 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2017/) · [**2016** 5 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2016/) · [**2015** 2 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2015/) · [**2014** 3 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2014/) · [**2013** 1 work](https://md.research.elamur.ai/free/en/papers/abrusan-2013-experimenting-king-france-topics-verifiability/) · [**2012** 1 work](https://md.research.elamur.ai/free/en/papers/clarke-2012-context-theoretic-framework-compositionality-distributional/)

### 2000s

[**2006** 1 work](https://md.research.elamur.ai/free/en/papers/koehn-2006-manual-automatic-evaluation-machine-translation/) · [**2004** 2 works](https://md.research.elamur.ai/free/en/agglomerations/london/year/2004/) · [**2000** 1 work](https://md.research.elamur.ai/free/en/papers/deemter-2000-coreferring-coreference-muc-related-annotation/)

### 1990s

[**1994** 1 work](https://md.research.elamur.ai/free/en/papers/lappin-1994-algorithm-pronominal-anaphora-resolution/)

## Concepts, shelves and topics

Of its works in our library: the number of works on each and its rating among them.

💡 Concepts 12[Language model · 3 works](https://md.research.elamur.ai/free/en/concepts/language-model/) · [Named entity recognition · 3 works](https://md.research.elamur.ai/free/en/concepts/named-entity-recognition/) · [Coreference resolution · 1 work](https://md.research.elamur.ai/free/en/concepts/coreference-resolution/) · [Knowledge representation and reasoning · 1 work](https://md.research.elamur.ai/free/en/concepts/knowledge-representation/) · [Semantic similarity · 1 work](https://md.research.elamur.ai/free/en/concepts/semantic-similarity/) · [BLEU score · 1 work](https://md.research.elamur.ai/free/en/concepts/bleu/) · [Lexical semantics · 1 work](https://md.research.elamur.ai/free/en/concepts/lexical-semantics/) · [Implicature · 2 works](https://md.research.elamur.ai/free/en/concepts/implicature/) · [Dependency parsing · 1 work](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/) · [Distributional semantics · 1 work](https://md.research.elamur.ai/free/en/concepts/distributional-semantics/) · [Psycholinguistics · 1 work](https://md.research.elamur.ai/free/en/concepts/psycholinguistics/) · [Cognitive linguistics · 1 work](https://md.research.elamur.ai/free/en/concepts/cognitive-linguistics/)

📚 Shelves 12[Natural language processing · 68 works](https://md.research.elamur.ai/free/en/shelves/nlp/) · [Other · 27 works](https://md.research.elamur.ai/free/en/shelves/other/) · [Linguistics & languages · 81 works](https://md.research.elamur.ai/free/en/shelves/linguistics/) · [Machine translation · 28 works](https://md.research.elamur.ai/free/en/shelves/machine-translation/) · [Mathematics & statistics · 7 works](https://md.research.elamur.ai/free/en/shelves/mathematics/) · [AI & machine learning · 10 works](https://md.research.elamur.ai/free/en/shelves/ai/) · [Computing & GPUs · 7 works](https://md.research.elamur.ai/free/en/shelves/computing/) · [English · 19 works](https://md.research.elamur.ai/free/en/shelves/en/) · [Life sciences & medicine · 1 work](https://md.research.elamur.ai/free/en/shelves/life-sciences/) · [Machine translation · 1 work](https://md.research.elamur.ai/free/en/shelves/mtrans/) · [Aramaic · 1 work](https://md.research.elamur.ai/free/en/shelves/arc/) · [Law · 1 work](https://md.research.elamur.ai/free/en/shelves/law/)

🧭 Topics 35[Language models & pretraining · 26 works](https://md.research.elamur.ai/free/en/topics/language-models/) · [Evaluation & benchmarks · 20 works](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) · [Multilingual & cross-lingual NLP · 21 works](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · [Machine translation · 21 works](https://md.research.elamur.ai/free/en/topics/machine-translation/) · [Coreference resolution · 22 works](https://md.research.elamur.ai/free/en/topics/coreference-resolution/) · [Question answering · 7 works](https://md.research.elamur.ai/free/en/topics/question-answering/) · [Sentiment & emotion · 14 works](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · [Reasoning · 7 works](https://md.research.elamur.ai/free/en/topics/reasoning/) · [Named entity recognition · 6 works](https://md.research.elamur.ai/free/en/topics/named-entity-recognition/) · [Code & programming languages · 3 works](https://md.research.elamur.ai/free/en/topics/code-and-programming/) · [Dialogue systems · 6 works](https://md.research.elamur.ai/free/en/topics/dialogue-systems/) · [Information retrieval & search · 3 works](https://md.research.elamur.ai/free/en/topics/information-retrieval/) · [Vision & language · 7 works](https://md.research.elamur.ai/free/en/topics/multimodal/) · [Text generation · 3 works](https://md.research.elamur.ai/free/en/topics/text-generation/) · [Information extraction · 2 works](https://md.research.elamur.ai/free/en/topics/information-extraction/)+20 more

Agglomerations: the GHSL Urban Centre Database 2025; populations: the GHSL functional urban areas 2015 (urban centre and commuting zone), else the urban centre, as each page says (European Commission, Joint Research Centre, CC BY 4.0). An institution belongs to the urban centre its coordinates (ROR, OpenAlex) fall in, or one within 5 km. Land: Natural Earth.

- [← back to](https://md.research.elamur.ai/free/en/countries/united-kingdom/) — United Kingdom

Updated 23 October 2026