Research map / Vision-language and multimodal / CLIP and zero-shot
OrdinalCLIP: Learning Rank Prompts for Language-Guided Ordinal Regression
Wanhua Li, Xiaoke Huang, Zheng Zhu, Yansong Tang, Xiu Li, Jie Zhou, Jiwen Lu
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 language prototypes with learned ordinal rank embeddings instead of a linear classification head | OrdinalCLIP · MORPH II | MAE: 2.63 -> 2.32 | Improved |
| Use language prototypes with learned ordinal rank embeddings instead of CoOp prompts | OrdinalCLIP · Adience | Accuracy: 60.6 ± 5.5 -> 61.2 ± 4.2 | Improved |
| Use language prototypes with learned ordinal rank embeddings instead of CoOp prompts | OrdinalCLIP · Image Aesthetics overall | MAE: 0.302 -> 0.280 | Improved |
| Use language prototypes with learned ordinal rank embeddings instead of CoOp prompts | OrdinalCLIP · Historical Image Dating | Accuracy: 51.90 ± 2.60 -> 56.44 ± 1.66 | Improved |
| Use 3 base ranks with linear interpolation rather than 2 base ranks (ablation) | OrdinalCLIP · MORPH II | MAE: 2.38 -> 2.32 | Improved |
| Use linear interpolation with 3 base ranks rather than inverse-proportion interpolation with 7 base ranks (ablation) | OrdinalCLIP · MORPH II | MAE: 2.35 -> 2.32 | Improved |
| Tune rank and context embeddings jointly rather than context embeddings only (ablation) | OrdinalCLIP · MORPH II | MAE: 2.37 -> 2.36 | Improved |
| Initialize context with the task-specific prompt and tune both embeddings (ablation) | OrdinalCLIP · MORPH II | MAE: 2.36 -> 2.32 | Improved |
| Tune rank embeddings only rather than both rank and context embeddings (ablation) | OrdinalCLIP · MORPH II | MAE: 2.32 -> 2.38 | Hurt |
| Tune context embeddings only rather than both rank and context embeddings (ablation) | OrdinalCLIP · MORPH II | MAE: 2.32 -> 2.37 | Hurt |
| Add ordinal rank embeddings to CoOp context-only tuning (ablation) | CoOp · MORPH II | MAE: 2.44 -> 2.39 | Improved |
| Use language-guided ordinal prompts instead of CNN Baseline in one-shot training | OrdinalCLIP · MORPH II, 1 shot | MAE: 8.04 -> 4.94 | Improved |
Later papers that compared with this method
3 later accepted papers report OrdinalCLIP as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- FluoCLIP: Stain-Aware Focus Quality Assessment in Fluorescence Microscopy (CVPR 2026 · better on 5 of 6 shared results)
- Interpretable Vision-Language Survival Analysis with Ordinal Inductive Bias for Computational Pathology (ICLR 2025 · better on 1 of 1 shared result)
- Teach CLIP to Develop a Number Sense for Ordinal Regression (ECCV 2024 · better on 3 of 3 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
- Ranking-aware adapter for text-driven image ordering with CLIP (ICLR 2025 · similarity 0.90)
- Learning to Prompt with Text Only Supervision for Vision-Language Models (AAAI 2025 · similarity 0.87)
- IPO: Interpretable Prompt Optimization for Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- Aggregate-and-Adapt Natural Language Prompts for Downstream Generalization of CLIP (NeurIPS 2024 · similarity 0.86)
- Meta CLIP 2: A Worldwide Scaling Recipe (NeurIPS 2025 · similarity 0.86)
- Probabilistic Language-Image Pre-Training (ICLR 2025 · similarity 0.86)
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.
