Research map / Label-efficient and robust learning / OOD and anomaly detection
Adaptive Prompt Learning via Gaussian Outlier Synthesis for Out-of-distribution Detection
Yongkang Zhang, Dongyu She, Zhong Zhou
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 location tokens to the CLIP-style prompt (ablation) | APLGOS prompts · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 51.31 / 50.96 -> 49.56 / 47.60 | Improved |
| initialize prompts from ChatGPT-sampled statements instead of "a region of a" (ablation) | APLGOS prompts · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 51.31 / 50.96 -> 49.50 / 49.40 | Improved |
| combine location tokens and sampled statements (full prompt) (ablation) | APLGOS prompts · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 49.56 / 47.60 -> 45.96 / 47.10 | Improved |
| use only class label as prompt vs full APLGOS prompt (ablation) | APLGOS prompts · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | mAP (ID): 48.2 -> 49.4 | Improved |
| APLGOS over VOS baseline (RegNetX4.0) (ablation) | VOS · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 50.53 / 50.27 -> 45.96 / 47.10 | Improved |
| APLGOS over VOS baseline (ResNet50) (ablation) | VOS · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 48.28 / 52.14 -> 47.16 / 49.66 | Improved |
| APLGOS over VOS (RegNetX4.0) on BDD100k (ablation) | VOS · Berkeley DeepDrive-100k (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 42.82 / 27.55 -> 39.48 / 19.79 | Improved |
| add Gaussian noise strength alpha=1.0 vs none (ablation) | APLGOS OOD prompt sampling · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 51.63 / 50.88 -> 45.96 / 47.10 | Improved |
| increase noise strength alpha from 1.0 to 1.5 (ablation) | APLGOS OOD prompt sampling · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 45.96 / 47.10 -> 55.88 / 53.33 | Hurt |
| increase ID:OOD ratio from 1:1 to 3:1 (ablation) | APLGOS training data · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | mAP (ID): 49.4 -> 47.5 | Hurt |
| increase ID:OOD ratio from 1:1 to 2:1 (baseline ratio) (ablation) | APLGOS training data · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95: 45.96 / 47.10 -> 48.25 / 50.20 | Hurt |
| vary number of sampled OOD prompts K around default 10000 (ablation) | APLGOS OOD prompt sampling · PASCAL VOC (ID) vs MS-COCO2017 / OpenImages (OOD) | FPR95 | Mixed |
Closest papers by meaning
- Local-Prompt: Extensible Local Prompts for Few-Shot Out-of-Distribution Detection (ICLR 2025 · similarity 0.91)
- Envisioning Outlier Exposure by Large Language Models for Out-of-Distribution Detection (ICML 2024 · similarity 0.90)
- ID-like Prompt Learning for Few-Shot Out-of-Distribution Detection (CVPR 2024 · similarity 0.90)
- Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection (ICML 2025 · similarity 0.90)
- Out-of-Distribution Detection with Prototypical Outlier Proxy (AAAI 2025 · similarity 0.88)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.88)
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.
