Research map / Label-efficient and robust learning / Self-supervised and contrastive
Unsupervised Semantic Segmentation Facilitates Model Understanding
Xiaoyan Yu, Lisa Mais, Peter Hirsch, Nick Lechtenbörger, Jannik Franzen, Andreas Mardt, Dagmar Kainmueller
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use token embeddings instead of key embeddings (ablation) | ViT SSL models · SPair-71k keypoint correspondence (ViT-B, best layer) | PCK@0.1: DINOv3 39.16 -> 57.87 | Improved |
| use token embeddings instead of key embeddings (ablation) | MAE ViT-B · SPair-71k keypoint correspondence (ViT-B, best layer) | PCK@0.1: 26.60 -> 19.48 | Hurt |
| increase positional effect in embeddings | ViT SSL models · COCO-Stuff-27 unsupervised segmentation | mIoU: rho = -0.59 | Hurt |
| stronger key/query positional effect | ViT SSL models · All layers | locality bias (I(Q,K)): rho = 0.70 | Improved |
| scale DINOv3 from Base to Large | DINOv3 · COCO-Stuff-27 unsupervised segmentation | mIoU / positional effect | Hurt |
| add register tokens to DINOv2 (ablation) | DINOv2 · COCO-Stuff-27 unsupervised segmentation | positional effect | Improved |
| scale iBOT from Base to Large | iBOT · COCO-Stuff animals | segmentation granularity | Improved |
| MIM (vs CL) pretraining | ViT-B key embeddings · COCO-Stuff-27 unsupervised segmentation | mIoU | Hurt |
Closest papers by meaning
- Causal Tracing of Object Representations in Large Vision Language Models: Mechanistic Interpretability and Hallucination Mitigation (AAAI 2026 · similarity 0.89)
- VideoHallu: Evaluating and Mitigating Multi-modal Hallucinations on Synthetic Video Understanding (NeurIPS 2025 · similarity 0.88)
- PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding (NeurIPS 2025 · similarity 0.88)
- To Sink or Not to Sink: Visual Information Pathways in Large Vision-Language Models (ICLR 2026 · similarity 0.88)
- What's in the Image? A Deep-Dive into the Vision of Vision Language Models (CVPR 2025 · similarity 0.88)
- UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
