Research map / 3D vision / 3D generation
Residual Relaxation for Multi-view Representation Learning
Yifei Wang, Zhengyang Geng, Feng Jiang, Chuming Li, Yisen Wang, Jiansheng Yang, Zhouchen Lin
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 |
|---|---|---|---|
| Add Prelax (residual relaxation + PL) to SimSiam | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 93.4 | Improved |
| Add Prelax to BYOL | BYOL ResNet-18 · ImageNette | accuracy (%): 91.9 -> 92.6 | Improved |
| Use Prelax-all instead of SimSiam | SimSiam · CIFAR-100 | accuracy (%): 66.9 -> 70.0 | Improved |
| Use Prelax-all instead of SimSiam | SimSiam · Tiny-ImageNet-200 | accuracy (%): 47.7 -> 49.2 | Improved |
| Use Prelax-rot instead of SimSiam | SimSiam · Tiny-ImageNet-200 | accuracy (%): 47.7 -> 47.1 | Hurt |
| Apply random rotation augmentation to each view in SimSiam (ablation) | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 87.9 | Hurt |
| Add margin loss relaxation (eta=0.5) to SimSiam (ablation) | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 91.9 | No clear effect |
| Add PL loss alone to SimSiam (ablation) | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 92.2 | Improved |
| Add R2S loss alone to SimSiam (no PL) (ablation) | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 91.5 | Hurt |
| Add RotPL loss alone to SimSiam (ablation) | SimSiam · CIFAR-10 | accuracy (%): 91.8 -> 91.1 | Hurt |
| Remove similarity constraint from Prelax-rot (ablation) | Prelax-rot · CIFAR-10 | accuracy (%): 92.4 -> 79.8 | Hurt |
| Use reverse residual r21 in R2S loss (ablation) | Prelax-std/all · CIFAR-10 | accuracy (%): 0.3 | Improved |
Closest papers by meaning
- ITO: Multi-View Alignment and Training-Time Fusion for Image-Text Pretraining (NeurIPS 2026 · similarity 0.87)
- Deep Incomplete Multi-view Learning via Cyclic Permutation of VAEs (ICLR 2025 · similarity 0.85)
- Rethinking Multi-view Representation Learning via Distilled Disentangling (CVPR 2024 · similarity 0.85)
- UVU: Improving Multimodal Understanding via Vision-Language Unified Autoregressive Paradigm (CVPR 2026 · similarity 0.85)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · similarity 0.85)
- Multi-View Causal Representation Learning with Partial Observability (ICLR 2024 · 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.
