Research map / Detection and segmentation / Open-set and few-shot detection
Representation Learning by Detecting Incorrect Location Embeddings
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add positional mismatch detection and randomized sparsity | MoCoV3 · ImageNet-1K | k-NN accuracy: 59.68 -> 64.69 | Helped |
| Add positional mismatch detection and randomized sparsity | DINO · ImageNet-1K | k-NN accuracy: 61.35 -> 63.74 | Helped |
| Add positional mismatch detection and randomized sparsity | SimCLR · ImageNet-1K | Linear_F accuracy: 50.21 -> 50.71 | Helped |
| Add DILEMMA to MoCoV3 pretraining | MoCoV3 · Yoga82 | accuracy: 56.41 -> 63.76 | Helped |
| Add DILEMMA to MoCoV3 pretraining | MoCoV3 · ADE20K, linear segmentation | mIoU: 12.44 -> 17.11 | Helped |
| Add DILEMMA loss to the MAE encoder | MAE · IN100 | Linear accuracy: 37.30 -> 39.06 | Helped |
| Use random instead of importance-based token dropping (ablation) | DILEMMA · IN100 | k-NN accuracy: 71.88 -> 73.98 | Helped |
| Randomize the token-dropping ratio instead of keeping it constant (ablation) | DILEMMA · IN100 | Linear accuracy: 77.78 -> 78.82 | Helped |
| Add mismatch detection to dense MoCoV3 inputs (ablation) | MoCoV3 · Yoga82 | Linear accuracy: 51.27 -> 56.15 | Helped |
| Add mismatch detection without sparsity (ablation) | MoCoV3 · ImageNet-1K | k-NN accuracy: 53.27 -> 54.18 | Helped |
| Increase the positional mismatch probability from 0.2 to 0.3 (ablation) | DILEMMA · ImageNet-1K | k-NN accuracy: 55.63 -> 55.34 | Hurt |
| Replace mismatch detection with position correction (ablation) | MoCoV3 · ImageNet-1K | k-NN accuracy: 55.63 -> 54.77 | Hurt |
Closest papers by meaning
- Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval (CVPR 2024 · similarity 0.86)
- Stochastic positional embeddings improve masked image modeling (ICML 2024 · similarity 0.86)
- Harnessing small projectors and multiple views for efficient vision pretraining (NeurIPS 2024 · similarity 0.86)
- Views Can Be Deceiving: Improved SSL Through Feature Space Augmentation (ICLR 2024 · similarity 0.85)
- Self-Supervised Learning of Intertwined Content and Positional Features for Object Detection (ICML 2025 · similarity 0.84)
- Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning (NeurIPS 2025 · similarity 0.84)
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.