Research map / Label-efficient and robust learning / Recognition and long-tail
Uncertainty Calibration for Ensemble-Based Debiasing Methods
Ruibin Xiong, Yimeng Chen, Liang Pang, Xueqi Cheng, Zhi-Ming Ma, Yanyan Lan
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 Dirichlet calibration of the bias-only model | PoE · FEVER Symmetric v1 | accuracy: 62.0 -> 65.9 | Improved |
| Add Dirichlet calibration of the bias-only model | PoE · FEVER Symmetric v2 | accuracy: 65.9 -> 69.1 | Improved |
| Add Dirichlet calibration of the bias-only model | Learned-Mixin · FEVER Symmetric v1 | accuracy: 59.8 -> 61.5 | Improved |
| Add Dirichlet calibration of the bias-only model | PoE · FEVER | in-distribution accuracy: 84.0 -> 87.1 | Improved |
| Add temperature scaling calibration of the bias-only model | PoE · FEVER Symmetric v1 | accuracy: 62.0 -> 63.3 | Improved |
| Use Dirichlet instead of temperature scaling calibrator | PoE · FEVER Symmetric v1 | accuracy: 63.3 -> 65.9 | Improved |
| Add Dirichlet calibration of the bias-only model | PoE · MNLI syntactic bias, HANS | accuracy: 68.1 -> 70.7 | Improved |
| Add Dirichlet calibration of the bias-only model | Learned-Mixin · MNLI syntactic bias, HANS | accuracy: 65.5 -> 62.7 | Hurt |
| Add Dirichlet calibration of the bias-only model | Learned-Mixin · MNLI hypothesis-only, Hard-SP | accuracy: 78.2 -> 83.2 | Improved |
| Add Dirichlet calibration of the bias-only model | PoE · MNLI hypothesis-only, Hard-CD | accuracy: 79.4 -> 79.4 | No clear effect |
| Add Dirichlet calibration of the bias-only model | PoE · MNLI hypothesis-only, in-distribution | accuracy: 83.2 -> 82.7 | Hurt |
| Apply Dirichlet calibration to bias-only model | bias-only model · FEVER | classwise ECE: 7.11 -> 1.73 | Improved |
Closest papers by meaning
- Denoising after Entropy-Based Debiasing a Robust Training Method for Dataset Bias with Noisy Labels (AAAI 2023 · similarity 0.86)
- Leveraging Uncertainty Estimates To Improve Classifier Performance (ICLR 2024 · similarity 0.85)
- SEBRA : Debiasing through Self-Guided Bias Ranking (ICLR 2025 · similarity 0.85)
- Mitigating Spurious Correlations via Disagreement Probability (NeurIPS 2024 · similarity 0.85)
- Improving Subgroup Robustness via Data Selection (NeurIPS 2024 · similarity 0.84)
- Truthful Calibration Errors for Multi-Class Prediction (NeurIPS 2026 · similarity 0.84)
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.
