Research map / Label-efficient and robust learning / OOD and anomaly detection
NegAS: Negative Label Guided Attention and Scoring for Out-of-Distribution Object Detection with Vision-Language Models
Yingjie Zhang, Shuai Li, Peng 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 |
|---|---|---|---|
| add NegA training and NegS scoring with LLM negative labels | YOLO-World · VOC vs OpenImages | FPR95: 48.8 -> 23.3 | Improved |
| add NegA training and NegS scoring | YOLO-World · VOC vs OpenImages | AUROC: 81.4 -> 92.6 | Improved |
| add negative-label guided attention and scoring on ID detection | YOLO-World + CoOp · VOC (ID detection) | mAP (ID): 72.2 -> 72.5 | No clear effect |
| generate negative labels with an LLM instead of lexical corpus mining (ablation) | NegAS · VOC vs OpenImages | FPR95: 42.4 -> 23.3 | Improved |
| generate negative labels with an LLM instead of lexical corpus mining (ablation) | NegAS · VOC vs MS-COCO | FPR95: 11.7 -> 9.9 | Improved |
| swap closed-source LLM for open-source Qwen3 to generate negative labels (ablation) | NegAS · VOC vs MS-COCO | FPR95: 9.9 -> 9.9 (AUROC 95.8) | No clear effect |
| add NegS scoring to the CoOp YOLO-World baseline (ablation) | YOLO-World + CoOp · VOC vs OOD sets | FPR95 / AUROC | Improved |
| add NegA on top of NegS (ablation) | YOLO-World + CoOp · VOC vs OOD sets | FPR95 / AUROC | Improved |
| add Gaussian noise to negative label embeddings (ablation) | NegAS · VOC vs OpenImages | FPR95 | Improved |
| restrict OOD loss to background regions (ablation) | NegAS · VOC vs OOD sets | OOD detection | Improved |
| set OOD loss coefficient alpha to 3 (ablation) | NegAS · VOC vs OOD sets | OOD detection | Improved |
Closest papers by meaning
- Negative Label Guided OOD Detection with Pretrained Vision-Language Models (ICLR 2024 · similarity 0.92)
- Conjugated Semantic Pool Improves OOD Detection with Pre-trained Vision-Language Models (NeurIPS 2024 · similarity 0.89)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.89)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- YolOOD: Utilizing Object Detection Concepts for Multi-Label Out-of-Distribution Detection (CVPR 2024 · similarity 0.89)
- Can We Ignore Labels in Out of Distribution Detection? (ICLR 2025 · 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.
