Get Started
Research questionHow can post-hoc OOD detection capture hierarchical features without sacrificing scale invariance or modifying the model?OOD signals may appear at different levels of a model’s representation, while those representations can have incompatible scales and noisy early-layer features. A useful detector must combine hierarchical information without losing reliable distance estimation.
AI
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.MM++: Post-Hoc Scale-Invariant Multilayer OOD Detection via Top-K Gated Feature FusionThe source presents MM++ for this setting, using selected intermediate and terminal features with a regularized tied covariance model. The reported evidence covers distinct architectures and both near- and far-OOD detection; no specific datasets or deployment constraints are provided.research paper · Sep 2, 2026
Related questions
How can training-free visual anomaly detection catch local defects and global rule violations without category-specific modeling?How can we detect object hallucinations in vision-language models when visual grounding shifts across layers?How can 3D class-incremental models learn new categories while remaining robust across heterogeneous point-cloud domains?How can vision systems discover unknown fine-grained classes without capture-device bias distorting cross-device evaluation?
Home
Topics
Search
Library