Research map / Vision-language and multimodal / CLIP and zero-shot
HyperCLIP++: Fine-Tuning CLIP for Open-Vocabulary Semantic Segmentation in Hyperbolic Space
Zelin Peng, Zhengqin Xu, Changsong Wen, Yu Huang, Yaoming Wang, Xiaokang Yang, Wei Shen
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 Mobius matrix multiplication (ablation) | HyperCLIP++ · A-847 | mIoU: 12.0 -> 13.1 | Improved |
| add Mobius matrix multiplication (ablation) | HyperCLIP++ · PC-459 | mIoU: 18.9 -> 20.3 | Improved |
| add dual cross-relation communication (DCRC) (ablation) | HyperCLIP++ · A-847 | mIoU: 12.3 -> 13.1 | Improved |
| add dual cross-relation communication (DCRC) (ablation) | HyperCLIP++ · PC-459 | mIoU: 19.2 -> 20.3 | Improved |
| use DCRC instead of cross-modality only communication (ablation) | HyperCLIP++ · A-847 | mIoU: 12.9 -> 13.1 | Improved |
| use DCRC instead of cross-layer only communication (ablation) | HyperCLIP++ · A-847 | mIoU: 12.5 -> 13.1 | Improved |
| use DCRC instead of cross-attention adapter (ablation) | HyperCLIP++ · A-847 | mIoU: 12.7 -> 13.1 | Improved |
| fine-tune both encoders instead of image encoder only (ablation) | HyperCLIP++ w/o DCRC · A-847 | mIoU: 10.3 -> 12.3 | Improved |
| fine-tune both encoders instead of text encoder only (ablation) | HyperCLIP++ w/o DCRC · A-847 | mIoU: 11.1 -> 12.3 | Improved |
| hyperbolic fine-tuning instead of Euclidean LoRA (ablation) | HyperCLIP++ · A-847 | mIoU: 11.4 -> 13.1 | Improved |
| increase diagonal block size n from 1 to 128 (ablation) | HyperCLIP++ · A-847 | mIoU: 6.0 -> 13.1 | Improved |
| increase diagonal block size n from 128 to 256 (ablation) | HyperCLIP++ · A-847 | mIoU: 13.1 -> 13.0 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CAT-Seg — CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Understanding Fine-tuning CLIP for Open-vocabulary Semantic Segmentation in Hyperbolic Space (CVPR 2025 · similarity 0.98)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.89)
- SuperCLIP: CLIP with Simple Classification Supervision (NeurIPS 2025 · similarity 0.89)
- Compositional Entailment Learning for Hyperbolic Vision-Language Models (ICLR 2025 · similarity 0.89)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.89)
- ResCLIP: Residual Attention for Training-free Dense Vision-language Inference (CVPR 2025 · similarity 0.89)
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.
