Closing Conference

About the Conference

After several years of collaborative research within the SAIL initiative, the SAIL Closing Conference brings together researchers from all partner institutions and the broader AI community to present results, exchange ideas and discuss future perspectives in sustainable and efficient AI.

Key Information

  • Date: Thursday, 7 May 2026
  • Location: CITEC Building, Bielefeld University
  • Target group: Researchers at all career stages, practitioners
  • Participation fee: None (free of charge)
  • Conference registration: https://indico.global/event/16794/overview

Program overview

Keynotes

Mira Mezini (TU Darmstadt)

AI-assisted Programming: From Intelligent Code Completion to Foundation Models. A Twenty-Year Journey

Abstract

From pioneering work on intelligent code completion to large language models, AI has have significant impact on software engineering over the past two decades. This talk traces the evolution of AI-assistedprogramming, highlighting advancements and outlining future directions. First, we’ll journey back to 2000-2010, briefly exploring pioneering applications of machine learning methods to coding tasks, in particular, the groundbreaking work from my lab on intelligent code completion, which was honored with the ACM SIGSOFT Impact Paper Award in 2024, showcasing the software engineering community’s early contributions. The second part of the talk examines the current landscape dominated by modern large language models (LLMs). Primarily driven by the ML community, these tools are being rapidly adapted by the software engineering community for various tasks. This part of the talk will highlight the pressing need for designing more reliable and specialized foundation models for software engineering tasks. Subsequently, I’ll present ongoing work from our lab focused on developing more robust foundation models for coding with the specific needs of software engineering in mind. This retrospective not only celebrates past achievements but also critically examines the present landscape, emphasizing the vital role of software engineering expertise in shaping the future of AI-assisted programming.

Nicola Strisciuglio (University of Twente)

Beyond Scaling: Toward Data-Efficient and Reliable Vision Models

Abstract

The recent success of computer vision foundation models has largely been driven by aggressive scaling of data, model size, and compute. While effective, this paradigm comes at a steep cost: high energy consumption, poorly controlled training data, embedded biases, and limited accessibility. From the perspective of sustainable computing, this trajectory is increasingly difficult to justify.
In this talk, I argue for a shift from brute-force learning: data efficiency and reliability as core research objectives.  will present empirical findings that reveal latent compositional structures in vision–language models, alongside persistent shortcut-driven biases that affect generalization abilities of current models. These observations motivate open research challenges: moving beyond post-hoc analysis toward training strategies that explicitly promote compositional representations and bias-aware learning. Building on work on prior knowledge in vision, I envision to lift priors, compositionality and bias-awareness into self-supervised learning objectives, embedding structure and similarity at a level that remains general, data-efficient, and scalable.
This perspective aims at a challenging prevailing scaling assumptions and opens new pathways toward vision foundation models that are easier to train and curate, and more transparent in what they learn. Ultimately, sustainability in AI is not only a hardware problem, but a learning problem: rethinking how models learn is key to building vision systems that are robust, efficient, and broadly accessible.

Katharina Eggensperger (TU Dortmund)

AutoML for Tabular Data

Junior Research Group Leader: Spotlights

Spotlights Part I

  • Michiel Straat: “Machine Learning for Control and Fast Prediction of Convective Flows” (Straat et al. 2026)
  • Andreas Besginow: “On Quick and Interpretable Selection of Physical Models through Gaussian Processes” (Besginow et al. 2026)
  • Michael Röder: “Explainable Benchmarking for Question Answering and Beyond” (Zhang et al. 2025)

Spotlights Part II

Poster Session

Poster_DigitalesDokumentDownload
Call for Abstracts

As part of the SAIL Closing Conference, we invite contributions to the Poster Session, which provides an informal and interactive setting for presenting research, exchanging ideas and receiving feedback. We welcome submissions from researchers at all career stages, including Master’s students, doctoral researchers and postdoctoral researchers. Posters may present:

  • completed research results
  • ongoing projects
  • innovative methods or applications
  • demonstrations of software, tools or systems
  • interdisciplinary work within or beyond the SAIL context

opics of Interest

We welcome poster submissions related to the broad research agenda of SAIL. Topics include, but are not limited to:

Sustainable & Efficient AI
  • resource-efficient model architectures and training methods
  • AI for sustainability (e.g. environmental modelling, smart energy systems)
  • hardware-aware algorithms and efficient computing
  • benchmarking, Green AI and lifecycle analysis
Human-Centered & Interactive AI
  • human–AI collaboration and interaction
  • explainable and interpretable AI
  • ethical, social and legal aspects of AI
  • AI for social good, healthcare and assistive technologies
Theoretical & Methodological Foundations
  • transfer, meta- and few-shot learning
  • learning from limited, imbalanced or noisy data
  • federated and privacy-preserving learning
  • robustness, optimization and generalization
Knowledge-Augmented & Neuro-Symbolic AI
  • hybrid approaches combining symbolic reasoning and neural models
  • knowledge graphs and structured knowledge integration
  • data-efficient learning through prior knowledge
Foundation Models & LLMs in Practice
  • efficient adaptation and fine-tuning of large language models
  • prompt engineering and in-context learning
  • applications and critical assessments of LLMs
AI for Modelling & Understanding Complex Systems
  • physics-informed and simulation-based machine learning
  • AI for control, optimization and dynamical systems
  • applications in climate science, biology and engineering

This list is indicative and not exhaustive. We aim to attract high-quality submissions from across the AI community and its numerous application fields. Whether your work presents a finished result, an ongoing project, or a compelling research idea, we invite you to share it with our interdisciplinary audience.

Submission Information

Submission Process

1. Abstract Submission2. Poster Upload
Please submit your poster title, a brief abstract (max. 300 words), and the names and affiliations of all authors via our online submission portal. The submission portal requires registration.Upon acceptance, presenters will be asked to upload a digital version of their final poster via the same system. Please note: All presenters are responsible for printing and bringing their physical poster to the conference.

Poster Format

  • Orientation: Portrait (vertical) preferred
  • Recommended size: A0 (84.1 × 118.9 cm)
  • Maximum width: 90 cm

Posters should be visually clear and easy to read. A logical structure, high-resolution graphics and sufficiently large fonts are recommended.

Dates

  • Abstract submission deadline for poster session:  15 March extended to March 23
  • Notification of decision for poster session: 22 March
  • Conference registration deadline for presenters: 15 April

How to Submit

➡️ Abstract submission:
Please submit your abstract through our online submission system:https://indico.global/event/16794/abstracts/. Information on poster upload will be sent together with the acceptance notification.

Organizing Committee