mlsanityData to 14 Oct 2026

What Does Animal Re-Identification Learn? Linear Biological Concepts and Their Origins in Visual Representations

arXiv 2609.06020 ↗First posted 5 Sep 2026cs.CV cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Conservation increasingly relies on camera traps that collect more wildlife imagery than experts can manually analyze, making animal re-identification (Re-ID) essential for monitoring individuals and populations. Yet understanding which cues drive model decisions is challenging for ViT-based Re-ID models, whose metric-learning objectives provide no explicit supervision for biological concepts. We ask whether such models nonetheless organize their representations along biologically meaningful axes. Using a DINOv3 backbone fine-tuned for Western lowland gorilla Re-ID with triplet-margin loss, we find that sex and age emerge as linear directions that generalize to held-out individuals, reaching up to 0.91 AUROC and being recoverable from a single image per individual. Activation steering further shows that the sex direction is causally used by the model, flipping a significant fraction of predictions to the opposite sex. Comparing off-the-shelf and fine-tuned backbones shows that Re-ID training does not create these concepts, but relocates them across the network. Finally, data attribution reveals that the representation we find reflects a graded biological axis, is redundantly encoded across the population and shaped by visually ambiguous individuals. Together, these findings show how interpretability can uncover both the biological structure and failure modes of Re-ID representations, providing a step toward auditable computer vision for wildlife monitoring.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 32 references that are in the map. First posted on arXiv 5 Sep 2026.

Builds on

22 of 32 references.

References · 22

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  2. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  3. 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,375
  4. 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
  5. 2016Understanding intermediate layers using linear classifier probescited by 1,353
  6. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 1,032
  7. 2017In Defense of the Triplet Loss for Person Re-Identificationcited by 786
  8. 2023The Linear Representation Hypothesis and the Geometry of Large Language Modelscited by 716
  9. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 712
  10. 2019Interpreting the Latent Space of GANs for Semantic Face Editingcited by 695
  11. 2021TransReID: Transformer-based Object Re-Identificationcited by 435
  12. 2023Language Models Represent Space and Timecited by 316
10 more
  1. 2023TRAK: Attributing Model Behavior at Scalecited by 283
  2. 2018Representer Point Selection for Explaining Deep Neural Networkscited by 187
  3. 2020Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?cited by 151
  4. 2024AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformerscited by 120
  5. 2022XAI for Transformers: Better Explanations through Conservative Propagationcited by 100
  6. 2019Multi-views Embedding for Cattle Re-identificationcited by 8
  7. 2024Sparse, Efficient and Explainable Data Attribution with DualXDAcited by 6
  8. 2025GorillaWatch: An Automated System for In-the-Wild Gorilla Re-Identification and Population Monitoringcited by 4
  9. 2023Attributing Learned Concepts in Neural Networks to Training Datacited by 3
  10. 2026Are We Recognizing the Jaguar or Its Background? A Diagnostic Framework for Jaguar Re-Identificationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024An Individual Identity-Driven Framework for Animal Re-Identification
  2. 2024Categorical Keypoint Positional Embedding for Robust Animal Re-Identification
  3. 2026Can Edge-Deployable Vision-Language Models Identify Species?
  4. 2025MetaWild: A Multimodal Dataset for Animal Re-Identification with Environmental Metadata
  5. 2026The Umwelt Representation Hypothesis: Rethinking Universality
  6. 2025Animal Re-Identification on Microcontrollers
  7. 2026Parameter-Efficient Vision-Language Adaptation with Continuous Metadata Conditioning for Animal Re-Identification
  8. 2026Advancing Wildlife Conservation through Multimodal Animal Re-Identification with Environmental Metadata