Programme

PDF versions of papers are linked below (see also proceedings for abstracts and full citation details).

Day 1

9:00 - 9:30: Registration

9:30 - 9:35: Chair's Welcome

9:35 - 10:25: Keynote

We Are On Fire: Increasing efficiency since 1776, Marloes de Valk

We are on fire is a short inquiry into the relation between computing, efficiency and the climate crisis, featuring a speedrun through the parallel histories of environmentalism, computing and communities of practice working on alternatives to a narrow focus on efficiency. These communities are not only reimagining the elements of (network) technology within their reach, they are also actively shaping the vocabulary describing their practice: permacomputing, computing within limits, collapse informatics, salvage computing, slow tech, redundant technology, low-tech, feminist technology, convivial computing, disability driven development and more. What initiatives already exist that provide examples and visions that bust the obfuscating myths of Big Tech and listen to environmental and community needs rather than those of share holders?

Marloes de Valk (NL) is a software artist and writer in the post-despair stage of coping with the threat of global warming and being spied on by the devices surrounding her. She completed a PhD at the Centre for the Study of the Networked Image at London South Bank University, in collaboration with The Photographer's Gallery, looking into how communities of practice are redefining technology on a damaged Earth.

10:25 - 10:45: Tea / Coffee Break

10:45 - 12:00: Session 1
Counting Computing's Carbon: Measurement, Models and Impact Assessment

Session chair: Abdessalam Elhabbash - Lancaster University

A Scoping Review of Methods to Measure the Energy and Carbon Footprint of Web Tracking and Advertising (Online presentation)
Nils Bonfils and Christoph Becker

The Ultimate Carbon Cost of a ChatGPT Query (Online presentation)
Paul Kron

Energy and CO₂ Footprint of Climate Model Intercomparison Projects (Online presentation)
Sergi Palomas

Monitoring Energy Usage of an AI Research Lab with an Interactive Dashboard (Online presentation)
Marta Adamska

Hot Games: A Holistic Assessment of the Planet Warming Emissions of Video Games Based on 2024–2025 Data (Online presentation)
Mike Hazas, Kevin C. Dalli, Arjun Menon and Ben Abraham

12:00 - 13:00: Lunch - (Vegetarian/Vegan)

13:00 - 14:00: Session 2
Making Sustainable Computing Happen: Software, Behaviour, and Institutional Change

Session chair: Wim Vanderbauwhede - University of Glasgow

When Do Microservices Save Energy? Evidence from Environmental Simulation Workflows (Online presentation)
Joshua Rowley and Abdessalam Elhabbash

Scoring and Gamification to Encourage Sustainable Use of Compute Clusters (In-person presentation)
Maximilian MacDonald, Chris McCaig, Sean MacAvaney, Matthew Barr and Lauritz Thamsen

Turning Interest into Institutional Change: Teaching Advocacy for Sustainable Research (Online presentation)
Kirsty Pringle, Lorna Smith, Erinma Ochu and Greg Wilson

Developing a Sustainable Sustainability Strategy for the STFC Scientific Computing Department (Online presentation)
Jessica Huntley and David McDonagh

14:00 - 14:10: Short Break

14:10 - 15:10 Session 3
Frugal, Circular, and Long-Lived Computing

Session chair: Lorna Smith - The University of Edinburgh

Estimating the Growth in Emissions from AI Data Centres (In-person presentation)
Wim Vanderbauwhede

Wasted Large Language Models: A Life Cycle Thinking Approach (Online presentation)
Erik Johannes Husom, Maria Emine Nylund and Ophelia Prillard

GPU Implementation of a Resource-Constrained Virtual Machine (In-person presentation)
Simone Li, Wim Vanderbauwhede, Vladislav Brusokas, Andrei Ghita and Shuxuan Li

Too Cheap to Matter: Over-Abundant Microchips, and What We Can Learn from Them (In-person presentation)
Adrian Friday, Fieke Jansen, Srinjoy Mitra and Gauthier Roussilhe

15:10 - 15:30: Tea / Coffee Break

15:30 - 16:15: Session 4
Sustainable AI Systems: Efficient and Carbon-Aware ML

Session chair: Christina Bremer - University of Cambridge

RequestRouter: Request-Boundary Routing for Efficient Single-GPU LLM Inference (Online presentation)
Aman Sunesh, Ali Alshehhi and Hivansh Dhakne

Could Model Partitioning Make Federated Learning More Sustainable? (In-person presentation)
Tobias Frohlich, Tiffany Vlaar and Lauritz Thamsen

The Energy Blind Spot: NVIDIA’s Flagship Edge AI Hardware Cannot Support Process-Level Energy Attribution (Online presentation)
Deepak Panigrahy and Aakash Tyagi

16:15 - 16:25: Short Break

16:25 - 17:30: Lightning Talks

Session chair: Paul Dempster - Lancaster University

Software-centric measures against premature hardware obsolescence (In-person presentation)
Stephen Kell

Small-scale solar-powered social media: power saving through planned downtime (Online presentation)
James Coxon

Carbon-Aware Workflow Scheduling Under Deadlines (In-person presentation)
Dominik Schweisgut

Towards Post-Growth AI: What Motivates its Usage? (Online presentation)
Lena Pohlmann, Sophie Lecheler, Hajo Boomgaarden and Emanuel Sallinger

Ambition or excess? Computational sufficiency in land surface modelling. (In-person presentation)
Joe Marsh Rossney, Carolynne Lord, Lily Gouldsbrough, Gordon Blair, Marcia Smith and Kelly Widdicks

Green TTA: Benchmarking the Energy Efficiency of Test-Time Adaptation Methods (Online presentation)
Vivian White and Evan Shelhamer Christina Bremer, Jan van der Scheer and Loïc Lannelongue

Does measuring carbon change behaviour? Designing a multi-centre trial in research organisations (In-person presentation)
Christina Bremer, Jan van der Scheer and Loïc Lannelongue

Adaptive Carbon-Aware Machine Learning Training under Uncertainty: A Unified Scheduling Framework (Online presentation)
Sufiyan Ul Rehman

A unified user-facing carbon and energy monitoring dashboard for Digital Research Infrastructures (In-person presentation)
Navirah Kamal, Laurent Gil, Andrew Turner and Loïc Lannelongue

Making transparent, verifiable sustainability claims on the web with carbon.txt and DIST (Online presentation)
Tim Cowlishaw, Hannah Smith, Chris Adams and Fershad Irani

Interoperability as a Carbon Reduction Strategy (In-person presentation)
Kern Tallett

Counting the Stones in my Computer : Unmaking the Chip Reality (Online presentation)
Coralie Gourguechon

Green BOA: Determining the environmental break-even point for ML-based data compression (In-person presentation)
Caterina Doglioni, Akshat Gupta, Thomas Elliott and Sanjiban Sengupta

Towards Sustainability-Aware Benchmarking and Optimisation of Generalised Matrix-Matrix Multiplication (GEMM) Operations for CPUs, GPUs, and FPGAs on HPC Systems (In-person presentation)
Ishaan Alidina and Sylvain Laizet

17:30 - 17:40: Call for Planning Future LOCOs

Session chair: Adrian Friday - Lancaster University

Dinner

Participants to make their own arrangements

Day 2

9:00 - 9:30: Registration and Coffee / Tea

9:30 - 11:30: Tutorial One (2 hours)
Circular AI Infrastructure: Repurposing Existing Compute for Scalable Workloads

Delivered by Kalavai (https://kalavai.net/)

As part of the LOCO 2026 programme at Lancaster University, this tutorial explores how AI infrastructure can be rethought through a circular, sustainability-driven lens. Rather than relying on continual hardware refresh cycles, participants will be introduced to an approach that aggregates underutilised, on-premise compute into a unified, high-performance environment.

Using the Kalavai platform, the session will demonstrate how existing resources can be repurposed to support modern AI workloads, including large language models and custom research pipelines. Attendees will gain practical insight into deploying and managing distributed compute across heterogeneous environments, while extending hardware lifetimes and reducing both cost and environmental impact. The tutorial is particularly relevant for researchers and practitioners seeking scalable AI solutions within constrained or sustainability-focused settings.

11:30 - 12:00: Tea / Coffee Break

12:00 - 13:30: Tutorial 2
Federated Learning with Flower: From Foundations to Sustainable Systems

Delivered by Flower AI (https://flower.ai/)

This extended tutorial provides a comprehensive introduction to federated learning within the context of LOCO 2026, combining theoretical foundations with hands-on system design. Participants will work through the core principles of decentralised model training, where data remains distributed across devices or domains, and models are trained collaboratively.

Using the Flower framework, attendees will explore key challenges including communication efficiency, scalability, and privacy-preserving mechanisms. The session also considers the broader system implications of federated approaches, including their impact on resource usage and environmental footprint. By the end of the tutorial, participants will have a clear understanding of how to design, implement, and evaluate federated learning systems that are both effective and sustainable for real-world deployment.

13:30-14:30: Lunch - (Vegetarian/Vegan)

14:30-16:00: Tutorial 2 – continues

16:00-16:30: Tea / Coffee Break, Wrap-up and close