![[2602.15712] Criteria-first, semantics-later: reproducible structure discovery in image-based sciences](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15712] Criteria-first, semantics-later: reproducible structure discovery in image-based sciences
Summary
This article presents a novel approach to structure discovery in image-based sciences, advocating for a criteria-first methodology that separates structure extraction from semantic labeling, enhancing reproducibility and adaptability in scientific analysis.
Why It Matters
The shift from a semantics-first to a criteria-first approach addresses limitations in current image analysis methods, particularly in dynamic and interdisciplinary contexts. This framework promotes reproducibility and flexibility, crucial for advancing research in rapidly evolving scientific fields.
Key Takeaways
- Proposes a criteria-first approach for structure discovery in image analysis.
- Separates structure extraction from semantic mapping to enhance reproducibility.
- Addresses challenges in open-ended scientific discovery and cross-domain comparability.
- Supports the notion of treating structural products as FAIR and AI-ready.
- Encourages plural interpretations of data without altering upstream extraction methods.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.15712 (cs) [Submitted on 17 Feb 2026] Title:Criteria-first, semantics-later: reproducible structure discovery in image-based sciences Authors:Jan Bumberger View a PDF of the paper titled Criteria-first, semantics-later: reproducible structure discovery in image-based sciences, by Jan Bumberger View PDF HTML (experimental) Abstract:Across the natural and life sciences, images have become a primary measurement modality, yet the dominant analytic paradigm remains semantics-first. Structure is recovered by predicting or enforcing domain-specific labels. This paradigm fails systematically under the conditions that make image-based science most valuable, including open-ended scientific discovery, cross-sensor and cross-site comparability, and long-term monitoring in which domain ontologies and associated label sets drift culturally, institutionally, and ecologically. A deductive inversion is proposed in the form of criteria-first and semantics-later. A unified framework for criteria-first structure discovery is introduced. It separates criterion-defined, semantics-free structure extraction from downstream semantic mapping into domain ontologies or vocabularies and provides a domain-general scaffold for reproducible analysis across image-based sciences. Reproducible science requires that the first analytic layer perform criterion-driven, semantics-free structure discovery, yielding stable partitions, structural...