Research map / Label-efficient and robust learning / OOD and anomaly detection
Leveraging Dark Knowledge for Intrinsic Multimodal Out-of-Distribution Detection
Nimeshika Udayangani Hewa Dehigahawattage, Sarah Erfani, Christopher Leckie
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 self multimodal OOD distillation | vanilla joint classifier · HMDB51 25/26 near-OOD | FPR95: 38.78 -> 34.99 | Improved |
| add self multimodal OOD distillation | vanilla joint classifier · UCF101 50/51 near-OOD | FPR95: 10.10 -> 5.92 | Improved |
| add distillation framework | Base · HMDB51 far-OOD | FPR95: 26.00 -> 21.88 | Improved |
| add distillation framework | A2D · HMDB51 far-OOD | FPR95: 20.58 -> 17.54 | Improved |
| add distillation framework | DPU · HMDB51 25/26 near-OOD | FPR95: 34.42 -> 33.25 | Improved |
| add distillation framework | DPU · UCF101 50/51 near-OOD | FPR95: 7.57 -> 7.84 | Hurt |
| add distillation framework | FM · HMDB51 far-OOD | AUROC: 94.12 -> 93.92 | No clear effect |
| use TCKD only (no NCKD) (ablation) | self-distillation · HMDB51 25/26 near-OOD | FPR95: 38.78 -> 34.99 | Improved |
| use NCKD only (no TCKD) (ablation) | self-distillation · HMDB51 far-OOD EPIC-Kitchens | FPR95: 22.12 -> 16.01 | Improved |
| weight modalities by likelihood instead of uniformly (ablation) | distillation framework · HMDB51 far-OOD | FPR95: 22.96 -> 21.88 | Improved |
| use all modalities as teachers instead of best modality only (ablation) | distillation framework · HMDB51 far-OOD | FPR95: 22.52 -> 21.88 | Improved |
| add logit distillation versus naive ensembling with feature fusion (ablation) | joint classifier · HMDB51 far-OOD | FPR95: 26.97 -> 21.88 | Improved |
Closest papers by meaning
- MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities (NeurIPS 2024 · similarity 0.92)
- Extremely Simple Multimodal Outlier Synthesis for Out-of-Distribution Detection and Segmentation (NeurIPS 2025 · similarity 0.90)
- DPU: Dynamic Prototype Updating for Multimodal Out-of-Distribution Detection (CVPR 2025 · similarity 0.89)
- Enhanced OoD Detection through Cross-Modal Alignment of Multi-Modal Representations (CVPR 2025 · similarity 0.89)
- RUNA: Object-Level Out-of-Distribution Detection via Regional Uncertainty Alignment of Multimodal Representations (AAAI 2025 · similarity 0.88)
- Strengthen Out-of-Distribution Detection Capability with Progressive Self-Knowledge Distillation (ICML 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.
