Research map / Label-efficient and robust learning / OOD and anomaly detection
Auxiliary Prompt Tuning of Vision-Language Models for Few-Shot Out-of-Distribution Detection
Wenjun Miao, Guansong Pang, Zihan Wang, Jin Zheng, Xiao Bai
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 auxiliary outlier data via APT to SCT | SCT (1-shot) · ImageNet-1k-OOD hard OOD | AUROC: 56.70 -> 65.62 | Improved |
| plug APT into LoCoOp and IDLike | LoCoOp, IDLike (1-shot) · ImageNet-1k standard OOD average | FPR95: 32.79 -> 28.02 (LoCoOp); 34.72 -> 31.05 (IDLike) | Improved |
| add background regularization from auxiliary data (Back) (ablation) | SCT (1-shot) · Standard OOD average | FPR95: 31.09 -> 29.02 | Improved |
| add foreground-background pair regularization (Pair) (ablation) | SCT (1-shot) · ImageNet-1k-OOD | AUROC: 56.70 -> 63.10 | Improved |
| add Pair regularization on top of Back (ablation) | SCT (1-shot) · ImageNet-1k-OOD | AUROC: 58.65 -> 65.62 | Improved |
| use Outlier Exposure (OE) with auxiliary data instead of APT (ablation) | SCT (1-shot) · Standard OOD average | FPR95: 31.09 -> 36.47 | Hurt |
| use EnergyOE with auxiliary data (ablation) | SCT (1-shot) · ImageNet-1k-OOD | FPR95: 94.06 -> 96.98 | Hurt |
| regularize foreground regions directly (Fore) without pairing (ablation) | SCT (1-shot) · Places365 and SUN scene OOD | FPR95 | Hurt |
| APT on Textures OOD set | SCT (1-shot) · Textures | FPR95: 48.87 -> 50.96 | Hurt |
| increase auxiliary data quantity relative to ID data (ablation) | APT · ImageNet-1K 1-shot | FPR95 | Improved |
| few-shot prompt tuning methods vs zero-shot on hard OOD | LoCoOp, IDLike, SCT · ImageNet-1k-OOD | AUROC: GL-MCM 60.32 vs LoCoOp 54.40 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- LoCoOp — LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning (NeurIPS 2023) · better on 2 of 2 shared results
- GalLoP — GalLop: Learning global and local prompts for vision-language models (ECCV 2024) · better on 1 of 1 shared result
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
- ID-like Prompt Learning for Few-Shot Out-of-Distribution Detection (CVPR 2024 · similarity 0.91)
- Local-Prompt: Extensible Local Prompts for Few-Shot Out-of-Distribution Detection (ICLR 2025 · similarity 0.91)
- Overcoming Shortcut Problem in VLM for Robust Out-of-Distribution Detection (CVPR 2025 · similarity 0.91)
- Self-Calibrated Tuning of Vision-Language Models for Out-of-Distribution Detection (NeurIPS 2024 · similarity 0.90)
- Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection (ICML 2025 · similarity 0.89)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · 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.
