GEODE uses per-sample cosine-similarity scaling in a norm loss to preserve feature geometry for universal scorer-compatible OOD detection, matching or exceeding OE performance on CIFAR benchmarks.
hub
Learning Confidence for Out-of-Distribution Detection in Neural Networks
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
abstract
Modern neural networks are very powerful predictive models, but they are often incapable of recognizing when their predictions may be wrong. Closely related to this is the task of out-of-distribution detection, where a network must determine whether or not an input is outside of the set on which it is expected to safely perform. To jointly address these issues, we propose a method of learning confidence estimates for neural networks that is simple to implement and produces intuitively interpretable outputs. We demonstrate that on the task of out-of-distribution detection, our technique surpasses recently proposed techniques which construct confidence based on the network's output distribution, without requiring any additional labels or access to out-of-distribution examples. Additionally, we address the problem of calibrating out-of-distribution detectors, where we demonstrate that misclassified in-distribution examples can be used as a proxy for out-of-distribution examples.
hub tools
citation-role summary
background 2
citation-polarity summary
verdicts
UNVERDICTED 12roles
background 2polarities
background 2representative citing papers
PROBE recasts MLIP uncertainty quantification as selective classification by training a compact discriminative classifier on frozen per-atom backbone embeddings, yielding a reliability probability that tracks actual error better than ensemble disagreement.
HealthPoint represents clinical events as points in a 4D space (content, time, modality, case) and applies low-rank relational attention to achieve state-of-the-art mortality prediction from multi-level incomplete multimodal EHRs.
Fold is a post-hoc OOD detector that exploits larger feature-Hessian curvature on OOD inputs together with partial feature normalization and a self-supervised AutoFold calibration scheme.
Introduces an architecture-agnostic Adapter Head and Invascal self-calibration objective to produce calibrated evidential uncertainty estimates for LiDAR range-view semantic segmentation while preserving accuracy.
PVLM combines parsing-aware vision-language modeling with dynamic contrastive learning to enable fine-grained zero-shot attribution of deepfakes to unseen generators and outperforms prior methods on a new benchmark.
CoOD decomposes inputs into components and applies Component Shift Score plus Compositional Consistency Score to improve detection of both standard and compositional out-of-distribution data.
Uncertainty optimization alone misses most safety gains; a decision-stage deferral policy removes up to 80% segmentation errors at 25% pixel deferral with cross-dataset robustness, while calibration does not improve decision quality.
RankOOD detects out-of-distribution samples by training a model to predict fixed class-specific ranking permutations via the Plackett-Luce loss, achieving a 4.3% FPR95 reduction on near-OOD TinyImageNet.
Benchmark across architectures and shift regimes finds OOD detector rankings shift with representation collapse; proposes NC-based shortlist predictor and PCA filter without extra OOD data.
An optimization-based deep learning pipeline selects informative patches from H&E whole-slide images to classify breast cancer into PAM50 subtypes, achieving F1 scores of 0.88 internally and 0.80 externally.
A multi-head RNN framework with learned confidence, ensemble uncertainty, auxiliary predictions, distance analysis, and diagnostics produces calibrated trust scores for NOx prediction, reducing MAE from 0.202 to 0.070 on the top 10% confidence subset.
citing papers explorer
-
GEODE: Angle-Adaptive OOD Detection with Universal Scorer Compatibility
GEODE uses per-sample cosine-similarity scaling in a norm loss to preserve feature geometry for universal scorer-compatible OOD detection, matching or exceeding OE performance on CIFAR benchmarks.
-
Knowing when to trust machine-learned interatomic potentials
PROBE recasts MLIP uncertainty quantification as selective classification by training a compact discriminative classifier on frozen per-atom backbone embeddings, yielding a reliability probability that tracks actual error better than ensemble disagreement.
-
A Clinical Point Cloud Paradigm for In-Hospital Mortality Prediction from Multi-Level Incomplete Multimodal EHRs
HealthPoint represents clinical events as points in a 4D space (content, time, modality, case) and applies low-rank relational attention to achieve state-of-the-art mortality prediction from multi-level incomplete multimodal EHRs.
-
Exploiting Local Flatness for Efficient Out-of-Distribution Detection
Fold is a post-hoc OOD detector that exploits larger feature-Hessian curvature on OOD inputs together with partial feature normalization and a self-supervised AutoFold calibration scheme.
-
Invascal: Inverse-Vacuity Self-Calibration for Uncertainty-Aware LiDAR Range-View Semantic Segmentation
Introduces an architecture-agnostic Adapter Head and Invascal self-calibration objective to produce calibrated evidential uncertainty estimates for LiDAR range-view semantic segmentation while preserving accuracy.
-
PVLM: Parsing-Aware Vision Language Model with Dynamic Contrastive Learning for Zero-Shot Deepfake Attribution
PVLM combines parsing-aware vision-language modeling with dynamic contrastive learning to enable fine-grained zero-shot attribution of deepfakes to unseen generators and outperforms prior methods on a new benchmark.
-
Component-Based Out-of-Distribution Detection
CoOD decomposes inputs into components and applies Component Shift Score plus Compositional Consistency Score to improve detection of both standard and compositional out-of-distribution data.
-
Rethinking Uncertainty in Segmentation: From Estimation to Decision
Uncertainty optimization alone misses most safety gains; a decision-stage deferral policy removes up to 80% segmentation errors at 25% pixel deferral with cross-dataset robustness, while calibration does not improve decision quality.
-
RankOOD -- Class Ranking-based Out-of-Distribution Detection
RankOOD detects out-of-distribution samples by training a model to predict fixed class-specific ranking permutations via the Plackett-Luce loss, achieving a 4.3% FPR95 reduction on near-OOD TinyImageNet.
-
A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts
Benchmark across architectures and shift regimes finds OOD detector rankings shift with representation collapse; proposes NC-based shortlist predictor and PCA filter without extra OOD data.
-
A deep learning pipeline for PAM50 subtype classification using histopathology images and multi-objective patch selection
An optimization-based deep learning pipeline selects informative patches from H&E whole-slide images to classify breast cancer into PAM50 subtypes, achieving F1 scores of 0.88 internally and 0.80 externally.
-
Trust-Aware Predictive Emissions Monitoring for Gas Turbine Fleets with Limited Labelled Data
A multi-head RNN framework with learned confidence, ensemble uncertainty, auxiliary predictions, distance analysis, and diagnostics produces calibrated trust scores for NOx prediction, reducing MAE from 0.202 to 0.070 on the top 10% confidence subset.