Research map / Label-efficient and robust learning / OOD and anomaly detection
FlowCon: Out-of-Distribution Detection using Flow-based Contrastive Learning
Saandeep Aathreya, Shaun Canavan
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 |
|---|---|---|---|
| train a flow model with Bhattacharyya-based flow contrastive loss on fixed ResNet18 penultimate features | FlowCon · CIFAR-10 far-OOD (ResNet18) | AUROC: 96.47 (Heatmap) -> 97.19 | Improved |
| FlowCon versus post-hoc methods in mixed near/far OOD | FlowCon · CIFAR-10 vs CIFAR-100 (ResNet18) | AUROC: 86.45 (MSP) -> 93.97 | Improved |
| FlowCon versus Energy in near-OOD with WideResNet | FlowCon · CIFAR-100 vs CIFAR-10 (WideResNet) | AUROC: 74.30 (Energy) -> 67.03 | Hurt |
| lower feature dimension (WideResNet, 128-d) instead of ResNet18 (512-d) | FlowCon · CIFAR-100 near-OOD | AUROC: 82.80 -> 67.03 | Hurt |
| use flow loss weight lambda 0.07 instead of larger or smaller values (ablation) | FlowCon · CIFAR-100 far-OOD (WideResNet) | AUROC: 75.62 (0.05), 78.57 (1.0) -> 83.62 (0.07) | Improved |
| classify with class-wise learned distributions (Bayes rule) instead of original classifier head | FlowCon · CIFAR-100 (ResNet18) | accuracy: 75.8 -> 74.9 | No clear effect |
| classify with FlowCon distributions on WideResNet | FlowCon · CIFAR-10 (WideResNet) | accuracy: 93.3 -> 93.8 | Improved |
| add contrastive loss to flow likelihood (class-specific multimodal latent) (ablation) | FlowCon · toy moons | likelihood separation | Improved |
Closest papers by meaning
- Understanding Likelihood of Normalizing Flow and Image Complexity through the Lens of Out-of-Distribution Detection (AAAI 2024 · similarity 0.89)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.86)
- ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection (ICLR 2024 · similarity 0.86)
- Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection (ICML 2025 · similarity 0.85)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.85)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.85)
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.
