Research map / Vision-language and multimodal / CLIP and zero-shot
Region-based Cluster Discrimination for Visual Representation Learning
Yin Xie, Kaicheng Yang, Xiang An, Kun Wu, Yongle Zhao, Weimo Deng, Zimin Ran, Yumeng Wang, Ziyong Feng, Roy Miles, Ismail Elezi, Jiankang Deng
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 |
|---|---|---|---|
| increase sampled boxes per image from 1 to 10 (ablation) | RICE · COCO-Det | AP: 31.4 -> 35.2 | Improved |
| increase sampled boxes from 10 to 20 (ablation) | RICE · COCO-Det | AP: 35.2 -> 34.7 | Hurt |
| increase cluster centers from 200K to 1M (ablation) | RICE · COCO-Det | AP: 32.8 -> 35.2 | Improved |
| increase cluster centers from 1M to 2M (ablation) | RICE · LVIS-Det | AP: 23.1 -> 24.0 | Improved |
| increase cluster centers from 1M to 2M (ablation) | RICE · COCO-Det | AP: 35.2 -> 34.9 | Hurt |
| lower negative sampling rate from 0.1 to 0.05 (ablation) | RICE · COCO-Det | AP: 35.2 -> 38.8 | Improved |
| raise negative sampling rate from 0.1 to 0.5 (ablation) | RICE · COCO-Det | AP: 35.2 -> 33.7 | Hurt |
| lower negative sampling rate from 0.1 to 0.05 (ablation) | RICE · COCO-Seg | AP: 28.6 -> 27.3 | Hurt |
| increase positive OCR labels from 3 to 10 (ablation) | RICE · OCRBench | score: 391 -> 403 | Improved |
| increase positive OCR labels from 10 to 20 (ablation) | RICE · OCRBench | score: 403 -> 401 | Hurt |
| use RICE instead of MLCD as vision encoder | MLCD · Roboflow100-VL average | AP: 17.3 -> 26.5 | Improved |
| use RICE instead of CLIP in LLaVA-NeXT | CLIP · OCRBench | score: 525 -> 575 | Improved |
Closest papers by meaning
- Locality Alignment Improves Vision-Language Models (ICLR 2025 · similarity 0.88)
- GranViT: A Fine-Grained Vision Model With Autoregressive Perception For MLLMs (ICLR 2026 · similarity 0.88)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.87)
- RegionGPT: Towards Region Understanding Vision Language Model (CVPR 2024 · similarity 0.87)
- Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: A Comprehensive Evaluation (ICLR 2026 · similarity 0.87)
- LLMs Meet VLMs: Boost Open Vocabulary Object Detection with Fine-grained Descriptors (ICLR 2024 · similarity 0.87)
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.
