Global Research PartnershipsCreate free account
Research Collaboration Index

Subject Award · Physical Sciences & Engineering · 2026

Computer Science: the ten finalists

The ten finalist universities are the ten highest-ranked on the Computer Science collaboration index, listed here alphabetically until the full index publishes on 15 September. Each finalist is represented by an academic whose recent work exemplifies why: a real project, drawn from the open scholarly record, cited so you can check it.

Finalists, not winners. The ten finalists in each subject are the ten highest-ranked universities on the published subject index; each is represented by an academic and a project identified from the open scholarly record, with the paper cited so every claim can be checked. The exact positions publish with the full index on 15 September 2026 at 09:00 UK, when the awards become a headline strand of the global launch. Winners are decided by editorial review and announced on 15 September 2026. The Subject Awards are exclusive to partner universities.

The finalists

Ten universities. Ten projects. Ten academics.

Mohammad Aladfaj

πŸ‡ΈπŸ‡¦ King Saud University · Co-author

Body-Worn Sensors for Recognizing Physical Sports Activities in Exergaming via Deep Learning Model

Physical games that get people moving are the starting point here, with body-worn sensors recognising sports activities through a deep learning model. Beyond fitness, the team present the system as multi-purpose: trained on a domain-specific dataset, the same gesture recognition and virtual reality depiction pipeline is proposed for other application areas.

IEEE Access · 2023 · 103 citations · co-authors across 3 countries

Zachari Swiecki

πŸ‡¦πŸ‡Ί Monash University · Corresponding author

Assessment in the age of artificial intelligence

Traditional assessment is hard to design, offers only discrete snapshots of performance, and often tests skills people now routinely use computers to perform. This paper sets out those problems, reviews artificial intelligence approaches that address them at least in part, and asks critically whether those approaches introduce fresh difficulties of their own for assessment practice.

Computers and Education Artificial Intelligence · 2022 · 368 citations

L. C. Kwek

πŸ‡ΈπŸ‡¬ Nanyang Technological University · Co-author

Noisy intermediate-scale quantum algorithms

Today's quantum computers are noisy, and this review asks what can usefully be done with them anyway. Surveying the noisy intermediate-scale era, it covers the simulation of many-body physics and chemistry, combinatorial optimisation and machine learning, and argues that the period has already produced new programming paradigms that later machines will build on.

Reviews of Modern Physics · 2022 · 1,767 citations · co-authors across 4 countries

Jinlan Fu

πŸ‡ΈπŸ‡¬ National University of Singapore · Co-author

Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing

Prompt-based learning reworks how language models are used: instead of training a model for each task, a textual prompt with unfilled slots lets a pretrained model fill in the answer. Surveying and organising research across this new paradigm in natural language processing, the work explains why the framework is so powerful and maps the emerging landscape for the field.

ACM Computing Surveys · 2022 · 3,825 citations · co-authors across 2 countries

Zihuai Lin

πŸ‡¦πŸ‡Ί The University of Sydney · Co-author

Federated Learning for Smart Healthcare: A Survey

Hospitals hold data that could improve care, but pooling it centrally is often impractical and raises real privacy concerns. This survey examines federated learning for smart healthcare, in which multiple institutions coordinate to train shared artificial intelligence models without exchanging raw data, covering the motivations, the requirements and the recent designs proposed for the field.

ACM Computing Surveys · 2022 · 789 citations · co-authors across 3 countries

Spiros Denaxas

πŸ‡¬πŸ‡§ University College London · Co-author

PROBAST+AI: an updated quality, risk of bias, and applicability assessment tool for prediction models using regression or artificial intelligence methods

Prediction models in healthcare increasingly rely on artificial intelligence, and the tools used to judge their trustworthiness have had to keep pace. The project delivers PROBAST+AI, an updated framework with targeted signalling questions covering both model development and model evaluation, helping users assess the quality, risk of bias and applicability of prediction models and the studies behind them.

BMJ · 2025 · 643 citations · co-authors across 10 countries

Yulin Chen

πŸ‡ΊπŸ‡Έ University of California, Berkeley · Co-author

Parameter-efficient fine-tuning of large-scale pre-trained language models

As language models grow, retraining and storing every parameter becomes prohibitively costly. This work reviews parameter-efficient adaptation, where only a small portion of a model's parameters is optimised while the rest stay fixed, and gathers the various designs under the single term 'delta-tuning', showing that very large models can be stimulated effectively by tuning very little.

Nature Machine Intelligence · 2023 · 1,004 citations

Zakhar Shumaylov

πŸ‡¬πŸ‡§ University of Cambridge · Corresponding author

AI models collapse when trained on recursively generated data

As generative models fill the internet with text, the next generation of models will inevitably train on their predecessors' output. This study shows that indiscriminate use of model-generated content in training causes irreversible defects, with the tails of the original distribution disappearing, an effect the authors call model collapse and demonstrate in language models, variational autoencoders and Gaussian mixture models.

Nature · 2024 · 697 citations · co-authors across 2 countries

Ping Luo

πŸ‡­πŸ‡° University of Hong Kong · Co-author

PVT v2: Improved baselines with pyramid vision transformer

Vision transformers have driven encouraging progress in computer vision, but computational cost limits their practical use. The work improves the pyramid vision transformer with a linear complexity attention layer, overlapping patch embeddings and a convolutional feed-forward network, cutting cost to linear scale and delivering strong results on classification, detection and segmentation, with the code released openly.

Computational Visual Media · 2022 · 2,353 citations · co-authors across 5 countries

Sebastian Farquhar

πŸ‡¬πŸ‡§ University of Oxford · Corresponding author

Detecting hallucinations in large language models using semantic entropy

Large language models can reason impressively yet still hallucinate false outputs, a flaw with consequences ranging from fabricated legal precedents to risks in medical settings. The study develops entropy-based uncertainty estimators that detect a subset of hallucinations called confabulations, offering a statistical method that works even on new questions to which humans might not know the answer.

Nature · 2024 · 783 citations

Exclusive to partners

The Subject Awards are open to partner universities.

Partners verify their data, feature their academics, and are eligible for the Subject Awards. Finalists are identified from open data; winners are decided by our editorial team and announced on 15 September 2026.

Explore partnership →

The wider slate

Fifteen more candidate partnerships behind the finalists.

πŸ‡ΊπŸ‡Έ Harvard University×πŸ‡ΊπŸ‡Έ Center for Astrophysics Harvard & Smithsonian

Performance of NIRCam on JWST in Flight

2023 · 372 citations · 962 joint works in field

πŸ‡­πŸ‡° Hong Kong Polytechnic University×πŸ‡¨πŸ‡³ Shenzhen University

Towards resilience in Industry 5.0: A decentralized autonomous manufacturing paradigm

2023 · 245 citations · 267 joint works in field

πŸ‡¨πŸ‡¦ University of Toronto×πŸ‡¨πŸ‡¦ University of Waterloo

Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models

2023 · 589 citations · 162 joint works in field

πŸ‡­πŸ‡° City University of Hong Kong×πŸ‡¨πŸ‡³ Harbin Institute of Technology

Towards Public Verifiable and Forward-Privacy Encrypted Search by Using Blockchain

2022 · 260 citations · 241 joint works in field

πŸ‡ΊπŸ‡Έ Stanford University×πŸ‡ΊπŸ‡Έ University of California, Berkeley

Efficient Memory Management for Large Language Model Serving with PagedAttention

2023 · 1,159 citations · 211 joint works in field

πŸ‡¦πŸ‡Ί The University of Queensland×πŸ‡¦πŸ‡Ί Queensland University of Technology

VarifocalNet: An IoU-aware Dense Object Detector

2021 · 1,048 citations · 301 joint works in field

πŸ‡­πŸ‡° Chinese University of Hong Kong×πŸ‡¨πŸ‡³ Shenzhen University

Systematic literature review on opportunities, challenges, and future research recommendations of artificial intelligence in education

2022 · 990 citations · 2,208 joint works in field

πŸ‡¨πŸ‡³ Shenzhen University×πŸ‡­πŸ‡° Chinese University of Hong Kong

Systematic literature review on opportunities, challenges, and future research recommendations of artificial intelligence in education

2022 · 990 citations · 2,208 joint works in field

πŸ‡¦πŸ‡Ί UNSW Sydney×πŸ‡¦πŸ‡Ί University of Canberra

A new distributed architecture for evaluating AI-based security systems at the edge: Network TON_IoT datasets

2021 · 632 citations · 446 joint works in field

πŸ‡¨πŸ‡­ ETH Zurich×πŸ‡¨πŸ‡­ University of Zurich

The Liver Tumor Segmentation Benchmark (LiTS)

2022 · 1,175 citations · 218 joint works in field

πŸ‡¦πŸ‡Ί University of Technology Sydney×πŸ‡¦πŸ‡Ί UNSW Sydney

Deep reinforcement learning in recommender systems: A survey and new perspectives

2023 · 174 citations · 258 joint works in field

πŸ‡ΊπŸ‡Έ New York University×πŸ‡ΊπŸ‡Έ Columbia University

Transfer Learning Under High-Dimensional Generalized Linear Models

2022 · 139 citations · 111 joint works in field

πŸ‡¬πŸ‡§ Imperial College London×πŸ‡¬πŸ‡§ University College London

Beyond Transmitting Bits: Context, Semantics, and Task-Oriented Communications

2022 · 654 citations · 157 joint works in field

πŸ‡©πŸ‡ͺ Technical University of Munich×πŸ‡©πŸ‡ͺ Ludwig-Maximilians-UniversitΓ€t MΓΌnchen

Generative AI

2023 · 1,231 citations · 450 joint works in field

πŸ‡ΊπŸ‡Έ Massachusetts Institute of Technology×πŸ‡ΊπŸ‡Έ Harvard University

Noisy intermediate-scale quantum algorithms

2022 · 1,680 citations · 459 joint works in field

← All subject awards · The full Computer Science index →