Research map / Vision-language and multimodal / CLIP and zero-shot
Feature calibration for unseen-class detection
Calibrates multimodal features to handle few-shot, unseen-class and out-of-distribution inputs.
50 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2022–2026) · an approach in CLIP and zero-shot, keyword cluster “prompt · tuning · pre trained”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
21 papers in 2023–24 and 27 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 5.7% to 5.4%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2022 · Predict, Prevent, and Evaluate: Disentangled Text-Driven Image Manipulation Empowered by Pre-Trained Vision-Language Model (CVPR 2022 · 39 citations)
- 2023 · Open-Set Fine-Grained Retrieval via Prompting Vision-Language Evaluator (CVPR 2023 · 29 citations)
- 2024 · Delving into Multimodal Prompting for Fine-Grained Visual Classification (AAAI 2024 · 70 citations)
- 2025 · Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · 22 citations)
- 2026 · CLIPPan: Adapting CLIP as a Supervisor for Unsupervised Pansharpening (AAAI 2026 · 3 citations)
Earliest papers here
- Predict, Prevent, and Evaluate: Disentangled Text-Driven Image Manipulation Empowered by Pre-Trained Vision-Language Model (CVPR 2022 · 39 citations)
- Transferring Pre-trained Multimodal Representations with Cross-modal Similarity Matching (NeurIPS 2022 · 3 citations)
- Open-Set Fine-Grained Retrieval via Prompting Vision-Language Evaluator (CVPR 2023 · 29 citations)
What has been tried
36 papers here report 435 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 6%, hurt 17%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add DaTP to the baseline | CUB-200-2011 · accuracy | 90.8 -> 91.3 | Helped | Delving into Multimodal Prompting for Fine-Grained Visual Classification (AAAI 2024 · 70 citations) |
| add the prototype-guided patch decoder instead of ABMIL aggregation | TIHD-RCC (16-shot), low-scale prompt · AUC | 76.8 ± 3.1 -> 79.4 ± 2.1 | Helped | ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · 52 citations) |
| add attentive fusion with retrieved names from external visual-name memory | COCO test · CIDEr | 134.4 -> 140.1 | Helped | EVCap: Retrieval-Augmented Image Captioning with External Visual-Name Memory for Open-World Comprehension (CVPR 2024 · 44 citations) |
| add PCO and feature rank loss | Cross-domain average · angular error (degrees) | 7.71 -> 6.97 | Helped | CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model (AAAI 2024 · 24 citations) |
| use asymmetric face-conditioned prompts {T0,T3} instead of fixed prompts {T0,T1} | Celeb-DF-v2, trained on FF++/DF · AUC (%) | 82.88 -> 88.41 | Helped | Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · 22 citations) |
| Add Sparse Mixture-of-Adapters instead of a single MLP-Adapter | CUHK-PEDES · R@1 | 71.00 -> 72.00 | Helped | DM-Adapter: Domain-Aware Mixture-of-Adapters for Text-Based Person Retrieval (AAAI 2025 · 19 citations) |
| add image-specific distribution (ISD) in prompt flow | MVTec-AD · image-level AUROC | 92.3 vs 89.1 without ISD | Helped | Bayesian Prompt Flow Learning for Zero-Shot Anomaly Detection (CVPR 2025 · 19 citations) |
| Replace GoPT with shallow VPT-style prompts | NYUDv2 · mIoU | 54.1 -> 50.2 | Hurt | Prompting Multi-Modal Image Segmentation with Semantic Grouping (AAAI 2024 · 17 citations) |
| add kernel-aware hierarchical graph (KAHG) | VisA 1-shot · AUROC | 89.7 -> 91.3 | Helped | Kernel-Aware Graph Prompt Learning for Few-Shot Anomaly Detection (AAAI 2025 · 16 citations) |
| replace Gaussian-noise pseudo-spoof features with prompt learning for live and spoof features | OCM→I · HTER (%) | 34.43 -> 26.64 | Helped | SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · 12 citations) |
| Add LAlign | CIFAR10, CIFAR+10, CIFAR+50, TinyImageNet · AUC | 66.7 -> 76.0; 81.3 -> 87.1; 79.8 -> 85.5; 58.1 -> 67.2 | Helped | Improving Open Set Recognition via Visual Prompts Distilled from Common-Sense Knowledge (AAAI 2024 · 10 citations) |
| add symmetry-related loss LS | MegaAge-Asian · CA(3) | 67.74 -> 69.60 | Helped | Learning Triangular Distribution in Visual World (CVPR 2024 · 7 citations) |
What reviewers will expect
Datasets and baselines named in the 16 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CIFAR-10 · 3 of 16 papers
- MVTecAD · 2 of 16 papers
- VisA · 2 of 16 papers
- ImageNet · 2 of 16 papers
- Food-101 · 2 of 16 papers
- Oxford-Pets · 2 of 16 papers
- CIFAR-100 · 2 of 16 papers
- CUB200-2011 · 2 of 16 papers
Baselines
- WinCLIP · 2 of 16 papers
- APRIL-GAN · 2 of 16 papers
- LaBo · 2 of 16 papers
- Grad-CAM · 2 of 16 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 48 papers; from OpenAlex and the papers' affiliations).
- Tencent · 3 papers
- Chinese Academy of Sciences · 3 papers
- Nanjing University of Science and Technology · 2 papers
- Tongji University · 2 papers
- University of Chinese Academy of Sciences · 2 papers
- Peking University · 2 papers
- Peng Cheng Laboratory · 2 papers
- Tsinghua University · 2 papers
- Nanjing University · 2 papers
- Huawei · 2 papers
Use it
Download BibTeX of 26 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Prompt learning for CLIP (102 papers)
- Test-time adaptation of VLMs (93 papers)
- Parameter-efficient visual tuning (41 papers)
Elsewhere in the topic
- Open-vocabulary alignment from VLMs (130 papers)
- Attribute-based zero-shot learning (115 papers)
- Adapting VLMs for zero-shot classification (108 papers)
- Recaptioning and text-supervised pretraining (73 papers)
- CLIP training recipes and data curation (72 papers)
- Language models for zero-shot reasoning (64 papers)
- Fine-grained improvements to CLIP (63 papers)
- Self-distillation for vision-language pretraining (48 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).