Abstract

As Artificial Intelligence systems grow in complexity and scale, so do their environmental and societal impact. This workshop explores the urgent need for rethinking AI in terms of sustainability, with a focus on ecological and environmental issues. Presentations and discussions will rely on two pillars: Sustainability in AI, focused on reducing the ecological footprint of machine learning research and deployment, and AI for Sustainability, highlighting applications to global environmental challenges. With an emphasis on introspection of current practices, the workshop aims to foster responsible innovation, shared metrics, and encourage the community to design methods that are efficient, accessible, and viable with constrained resources.

Schedule (not final)

Time		Speaker
08:30	Welcome	–
08:40	Speaker 1	Loïc Lannelongue
09:35	Speaker 2	Claire Monteleoni
10:30	☕ Coffee Break	–
11:00	Poster Spotlights	–
11:30	Poster Session	–
12:30	🍽️ Lunch	–
13:30	Speaker 3	Bernd Ensing
14:00	Speaker 4	Jan-Willem van de Meent
14:30	Speaker 5	Sina Samangooei
15:00	☕ Coffee Break	–
15:45	🎤 Panel Discussion	All the speakers
16:45	Closing remarks	–

Accepted papers

📑 Accepted Papers

Type	Title	Authors
Short	Approximate Matrix Multiplication for Energy-Efficient Training of LLMs	Nils Kasper, Anna Kazachkova, Rainer Schlosser, Ralf Herbrich
Short	Sustainable AI research? Researchers’ perceptions of the environmental impact of AI and emissions-feedback tools – a mixed methods study	Theresa Willem, Marie Piraud
Long	xLSTM Scaling Laws: Competitive Performance with Linear Time-Complexity	Maximilian Beck, Kajetan Schweighofer, Sebastian Böck, Sebastian Lehner, Sepp Hochreiter
Short	JumpLM - LLM Benchmarking and Interactive Performance Monitoring for higher GPU Utilization	Lena Jurkschat, Anton Rygin, Elias Werner
Short	NA-LR: Noise-Adaptive Low-Rank Parameterisation for Efficient Diffusion Models	Jingyuan Wang, Federico Ottomano, Yingzhen Li
Long	Cut Less, Fold More: Model Compression through the Lens of Projection Geometry	Olga Saukh, Dong Wang, Haris Šikić, Yun Cheng, Lothar Thiele
Long	Quant-Trim in Practice: Improved Cross-Platform Low-Bit Deployment on Edge NPUs	Rayen Dhahri, Steffen Urban
Long	Energy-Aware Benchmarking: A Case Study on Sampling Methods	Anna Kazachkova, Johann Ukrow, Sven Köhler, Nicolas Alder, Rainer Schlosser, Ralf Herbrich
Long	KaVa: Latent Reasoning via Compressed KV-Cache Distillation	Anna Kuzina, Maciej Pióro, Paul N. Whatmough, Babak Ehteshami Bejnordi
Short	Position: From Abundance to Frugality. Why the Future of Frontier AI Depends on Data Reduction	Sophia N. Wilson, Raghavendra Selvan, Sebastian Mair
Long	Climate And Resource Awareness is Imperative to Achieving Sustainable AI (and Preventing a Global AI Arms Race)	Pedram Bakhtiarifard, Pınar Tözün, Christian Igel, Raghavendra Selvan
Long	STaMP: Sequence Trasformation and Mixed Precision for Low-Precision Activation Quantization	Marco Federici, Riccardo Del Chiaro, Boris van Breugel, Markus Nagel, Paul N. Whatmough
Long	xLSTM Distillation: Achieving Teacher-Student Parity Through Efficient Hybrid Architectures	Lukas Hauzenberger, Niklas Schmidinger, Thomas Schmied, Anamaria-Roberta Hartl, David Stap, Pieter-Jan Hoedt, Sebastian Böck, Günter Klambauer, Sepp Hochreiter
Long	Optimizing Large Language Models: Metrics, Energy Efficiency, and Case Study Insights	Tahniat Khan, Soroor Motie, Sedef Akinli Kocak, Shaina Raza
Short	The impacts of AI on environmental sustainability and human well-being	Noemi Luna Carmeno, Daniel Walton O’Neill, Tiago Domingos
Short	Efficiency Will Not Lead to Sustainable Reasoning AI	Philipp Wiesner, Daniel O’Neill, Francesca Larosa, Odej Kao
Short	Modelling the Doughnut of social and planetary boundaries with frugal machine learning	Stefano Vrizzi, Daniel Walton O’Neill
Long	SpikeFit: Towards Optimal Deployment of Spiking Networks on Neuromorphic Hardware	Ivan Kartashov, Mariia Pushkareva, Iakov Karandashev
Long	Energy Scaling Laws for Diffusion Models: Quantifying Compute and Carbon Emissions in Image Generation	Aniketh Iyengar, Jiaqi Han, Boris Ruf, Vincent Grari, Marcin Detyniecki, Stefano Ermon
Long	CacheSaver: A Modular Framework for Efficient, Affordable, and Reproducible LLM Inference	Nearchos Potamitis, Lars Henning Klein, Bardia Mohammadi, Chongyang Xu, Attreyee Mukherjee, Laurent Bindschaedler, Niket Tandon, Akhil Arora
Long	TiME: Tiny Monolingual Encoders for Efficient NLP Pipelines	David Schulmeister, Valentin Hartmann, Lars Henning Klein, Robert West

Important Dates

Submission Deadline: ~~October 17th, 2025, 23:59 AoE~~
Accept / Reject notification: ~~October 31st, 2025~~

Submission link

Submissions will be managed through OpenReview:

~~https://openreview.net/group?id=EurIPS.cc/2025/Workshop/RethinkingAI~~ (Submissions are closed)
Note that newly created profiles on OpenReview without an institutional email will go through moderation process that can take up to two weeks.

Call for Voluntary Reviewers

We are currently seeking voluntary reviewers to assist in the evaluation of submissions. If you have relevant expertise and are willing to contribute your time and insights, please fill out our reviewer sign-up form: ~~https://forms.gle/sVbzN9zs2UqyVc2n6~~
Reviewers must have an OpenReview account to participate. Your involvement will play a crucial role in ensuring the quality of the work presented at the workshop.

Call for Contributions

We invite submissions to the Rethinking AI workshop, which will bring together researchers, practitioners, and policymakers to discuss the dual challenges of applying AI for environmental preservation and making AI itself more environmentally responsible.

The workshop will feature two complementary tracks:

Track 1: Sustainability in AI

We aim to discuss the environmental footprint of AI itself. We encourage submissions that address:

Measuring and reporting the carbon, energy, and water footprint of AI models
Algorithmic and hardware efficiency (compression, pruning, quantization, frugal AI)
Lifecycle impacts of AI infrastructure (training, deployment, hardware production and disposal)
Socio-technical and governance perspectives on responsible AI development
Position papers reflecting on current research norms and proposing strategies for sustainable AI practices

Track 2: AI for Sustainability

We also welcome work that demonstrates how AI can contribute to tackling urgent environmental challenges. Submissions may include, but are not limited to, applications of machine learning for:

Climate science and modeling
Biodiversity monitoring and conservation
Agriculture and food systems
Renewable energy and grid optimization
Sustainable urban planning and mobility
Resource management and the circular economy

Submission Guidelines

We invite submissions of novel and unpublished work, as either short papers of at least 2 pages and up to 4 pages, or long papers of up to 9 pages. All page limits are excluding references and appendix. Submissions should follow NeurIPS formatting style, the checklist can be optionally left out.
Published papers can also be submitted in their original format, with the reviews included as supplementary material.
We welcome research papers, position papers, and case studies. Accepted contributions are non-archival (not published in proceedings) and will be presented as posters, with selected works invited for short spotlight talks. We expect at least one author to present at the workshop.