Research map / Label-efficient and robust learning / OOD and anomaly detection
FA: Forced Prompt Learning of Vision-Language Models for Out-of-Distribution Detection
Xinhua Lu, Runhe Lai, Yanqi Wu, Kanghao Chen, Wei-Shi Zheng, Ruixuan Wang
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 forced cross-entropy loss FCE-K instead of standard CE (ablation) | FA · ImageNet-1k OOD (16-shot) GL-MCM | FPR95: 34.29 -> 25.68 | Improved |
| use forced cross-entropy loss FCE-K instead of standard CE (ablation) | FA · ImageNet-1k OOD (16-shot) MCM | AUROC: 91.01 -> 93.77 | Improved |
| add manual initialization to the original (reference) prompt (ablation) | FA · ImageNet-1k OOD (16-shot) GL-MCM | FPR95: 34.88 -> 26.39 | Improved |
| add manual initialization to the forced prompt on top of manual original prompt (ablation) | FA · ImageNet-1k OOD (16-shot) GL-MCM | FPR95: 26.39 -> 25.68 | Improved |
| manually initialize only the forced prompt (original prompt random) (ablation) | FA · ImageNet-1k OOD (16-shot) GL-MCM | FPR95: 34.88 -> 36.03 | Hurt |
| use shared learnable vector instead of per-class independent vectors (random init) (ablation) | FA · ImageNet-1k OOD (16-shot) GL-MCM | FPR95: 46.29 -> 34.88 | Improved |
| use shared learnable vector with manual init (ablation) | FA · ImageNet-1k OOD (16-shot) MCM | FPR95: 34.69 -> 29.07 | Improved |
| increase forced coefficient K from 1 to 6 (ablation) | FA · ImageNet-1k OOD | AUROC | No clear effect |
| FA with GL-MCM vs SCT_GL baseline | FA_GL · ImageNet-1k OOD (16-shot) | FPR95: 27.27 -> 25.68 | Improved |
| FA_MCM vs SCT_GL baseline in 1-shot | FA_MCM · ImageNet-1k OOD (1-shot) | FPR95: 31.62 -> 31.37 | Improved |
| FA on SUN OOD set in 16-shot vs SCT_GL | FA_GL · ImageNet-1k OOD SUN (16-shot) | FPR95: 20.00 -> 27.65 | Hurt |
| FA_MCM vs LoCoOp_GL on EuroSAT ID | FA_MCM · EuroSAT as ID (16-shot) | FPR95: 78.69 -> 39.99 | Improved |
Closest papers by meaning
- Local-Prompt: Extensible Local Prompts for Few-Shot Out-of-Distribution Detection (ICLR 2025 · similarity 0.90)
- ID-like Prompt Learning for Few-Shot Out-of-Distribution Detection (CVPR 2024 · similarity 0.89)
- Conjugated Semantic Pool Improves OOD Detection with Pre-trained Vision-Language Models (NeurIPS 2024 · similarity 0.89)
- Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection (ICML 2025 · similarity 0.89)
- Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization (ICLR 2024 · similarity 0.89)
- Overcoming Shortcut Problem in VLM for Robust Out-of-Distribution Detection (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.
