Research map / Label-efficient and robust learning / OOD and anomaly detection
ReAct: Out-of-distribution Detection With Rectified Activations
Yiyou Sun, Chuan Guo, Yixuan Li
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 activation clipping at 90th percentile (ReAct) to energy score | ResNet-50 energy score · ImageNet-1k OOD avg | FPR95: 58.41 -> 31.43 | Improved |
| add ReAct to energy score | ResNet-18 CIFAR-10 · CIFAR-10 OOD avg | FPR95: 35.60 -> 32.91 | Improved |
| add ReAct to energy score | ResNet-18 CIFAR-100 · CIFAR-100 OOD avg | FPR95: 71.93 -> 59.61 | Improved |
| add ReAct to softmax score | ResNet-50 MSP · ImageNet-1k OOD avg | FPR95: 66.95 -> 58.28 | Improved |
| add ReAct to ODIN | ResNet-50 ODIN · ImageNet-1k OOD avg | FPR95: 56.48 -> 44.10 | Improved |
| lower clipping percentile from 90 to 65 (ablation) | ReAct ResNet-50 · ImageNet-1k OOD avg | FPR95: 31.43 -> 74.62 | Hurt |
| raise clipping percentile from 90 to 99 (ablation) | ReAct ResNet-50 · ImageNet-1k OOD avg | FPR95: 31.43 -> 44.57 | Hurt |
| clip at p=90 instead of no clipping (ablation) | ResNet-50 · ImageNet-1k val | ID accuracy: 75.08 -> 73.75 | Hurt |
| apply ReAct at penultimate layer instead of layer 3 (ablation) | ResNet-50 · ImageNet-1k OOD avg | FPR95: 63.87 -> 31.43 | Improved |
| add ReAct to GroupNorm model (ablation) | ResNet-50 GroupNorm · ImageNet-1k OOD avg | FPR95: 66.28 -> 51.58 | Improved |
| add ReAct to WeightNorm model (ablation) | ResNet-50 WeightNorm · ImageNet-1k OOD avg | FPR95: 50.34 -> 26.80 | Improved |
Later papers that compared with this method
28 later accepted papers report ReAct as a baseline in their results tables; 28 of them report a better number on most of the results they share with it.
- Ranked Activation Shift for Post-hoc Out-of-Distribution Detection (ECCV 2026 · better on 4 of 4 shared results)
- Enhancing Out-of-Distribution Detection with Extended Logit Normalization (CVPR 2026 · better on 2 of 3 shared results)
- Sparsity as a Key: Unlocking New Insights from Latent Structures for Out-of-Distribution Detection (CVPR 2026 · better on 2 of 2 shared results)
- Exploiting Local Flatness for Efficient Out-of-Distribution Detection (ECCV 2026 · better on 2 of 2 shared results)
- AdaSCALE: Adaptive Scaling for OOD Detection (ICML 2026 · better on 1 of 1 shared result)
- Gradient Short-Circuit: Efficient Out-of-Distribution Detection via Feature Intervention (ICCV 2025 · better on 2 of 2 shared results)
- Large Self-Supervised Models Bridge the Gap in Domain Adaptive Object Detection (CVPR 2025 · better on 1 of 1 shared result)
- A Variational Information Theoretic Approach to Out-of-Distribution Detection (ICML 2025 · better on 1 of 1 shared result)
20 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Detecting Out-of-Distribution Through the Lens of Neural Collapse (CVPR 2025 · similarity 0.89)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · similarity 0.89)
- Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement (ICLR 2024 · similarity 0.88)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.88)
- NECO: NEural Collapse Based Out-of-distribution detection (ICLR 2024 · similarity 0.88)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · 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.
