Research map / Label-efficient and robust learning / Recognition and long-tail
Trustworthy Classification for Complex Social Surveys: A Memory-Enhanced Hierarchical Framework with Calibrated Uncertainty
Zeqiang Wang, Rebecca Oldroyd, Yuqi Wang, Jiageng Wu, Jie Yang, Wei Wang, Nishanth R. Sastry, Jon Johnson, Suparna De
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 HOCM to BERT | BERT · Social survey in-dist test | Accuracy (%): 85.3 -> 84.9 | Hurt |
| add HOCM to BERT, evaluate on confident subset | BERT · Social survey in-dist test | Corrected accuracy (%): 84.9 -> 92.1 | Improved |
| add HOCM to BERT | BERT · Social survey open-set detection | MVR (%): 68.0 -> 78.9 | Improved |
| HOCM vs dedicated OSR methods (LOS) | BERT · Social survey open-set detection | MVR (%): 78.8 -> 78.9 | No clear effect |
| add HOCM to CharCNN+BiLSTM | CharCNN+BiLSTM · Social survey in-dist test | Accuracy (%): 77.1 -> 80.7 | Improved |
| add memory-enhanced contrastive learning (ablation) | BERT + hierarchical UC · Social survey ablation | Corrected accuracy (%): 89.0 -> 92.1 | Improved |
| add memory-enhanced contrastive learning (ablation) | BERT + hierarchical UC · Social survey ablation | MVR (%): 74.5 -> 78.9 | Improved |
| add quality filter in memory management (ablation) | HOCM · Social survey ablation | MVR (%): 78.0 -> 78.9 | Improved |
| add reliability-guided sampling (ablation) | HOCM · Social survey ablation | MVR (%): 78.2 -> 78.9 | Improved |
| use hierarchical calibration instead of flat OpenMax (ablation) | HOCM · Social survey ablation | MVR (%): 76.5 -> 78.9 | Improved |
| add hierarchical uncertainty calibration (ablation) | BERT + Memory CL · Social survey ablation | MVR (%): 72.5 -> 78.9 | Improved |
| add memory CL on closed-set accuracy (ablation) | BERT + hierarchical UC · Social survey ablation | Accuracy (%): 85.1 -> 84.9 | Hurt |
Closest papers by meaning
- Trusted Multi-view Learning for Long-tailed Classification (AAAI 2026 · similarity 0.83)
- Navigating Conflicting Views: Harnessing Trust for Learning (ICML 2025 · similarity 0.81)
- Prompts to Proxies: Emulating Human Preferences via a Compact LLM Ensemble (NeurIPS 2026 · similarity 0.81)
- Uncertainty Quantification for LLM-Based Survey Simulations (ICML 2025 · similarity 0.80)
- Value Imprint: A Technique for Auditing the Human Values Embedded in RLHF Datasets (NeurIPS 2024 · similarity 0.80)
- CORAL: Disentangling Latent Representations in Long-Tailed Diffusion (NeurIPS 2025 · similarity 0.80)
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.
