Research map / Label-efficient and robust learning / Recognition and long-tail
Learning to Predict Trustworthiness with Steep Slope Loss
Yan Luo, Yongkang Wong, Mohan Kankanhalli, Qi Zhao
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 |
|---|---|---|---|
| Replace TCP confidence loss with steep slope loss | Trustworthiness oracle with ViT backbone and ViT classifier · ImageNet val | AUC: 60.14±0.47 -> 73.68±0.27 | Improved |
| Replace TCP confidence loss with steep slope loss | Trustworthiness oracle with ViT backbone and ResNet classifier · ImageNet val | AUC: 66.23±1.02 -> 74.31±0.21 | Improved |
| Replace TCP confidence loss with steep slope loss | Trustworthiness oracle with ResNet backbone and ViT classifier · ImageNet val | AUC: 55.55±0.46 -> 64.34±0.16 | Improved |
| Replace TCP confidence loss with steep slope loss | Trustworthiness oracle with ResNet backbone and ResNet classifier · ImageNet val | AUC: 62.73±0.14 -> 67.92±0.11 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · Stylized ImageNet val | AUC: 57.29 -> 67.68 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · Adversarial ImageNet val | AUC: 55.81 -> 65.36 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · ImageNet-C Defocus blur | AUC: 54.27 -> 64.34 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · ImageNet-C Glass blur | AUC: 58.78 -> 70.18 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · ImageNet-C Motion blur | AUC: 58.36 -> 69.52 | Improved |
| Replace TCP confidence loss with steep slope loss | ViT trustworthiness oracle · ImageNet-C Zoom blur | AUC: 62.47 -> 70.94 | Improved |
| Replace TCP† fine-tuning loss with steep slope loss | Trustworthiness predictor · MNIST | FPR-95%-TPR: 3.33 -> 2.22 | Improved |
| Replace TCP† fine-tuning loss with steep slope loss | Trustworthiness predictor · CIFAR-10 | FPR-95%-TPR: 45.07 -> 44.69 | Improved |
Closest papers by meaning
- Certified Trustworthiness in the Era of Large Language Models (AAAI 2025 · similarity 0.87)
- Navigating Conflicting Views: Harnessing Trust for Learning (ICML 2025 · similarity 0.84)
- Towards Trustworthy Deep Learning (AAAI 2024 · similarity 0.83)
- U-trustworthy Models. Reliability, Competence, and Confidence in Decision-Making (AAAI 2024 · similarity 0.82)
- TRAM: Bridging Trust Regions and Sharpness Aware Minimization (ICLR 2024 · similarity 0.82)
- Dist Loss: Enhancing Regression in Few-Shot Region through Distribution Distance Constraint (ICLR 2025 · similarity 0.82)
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.
