Research map / Large language models / Language, safety and interpretability
Self-explaining deep models with logic rule reasoning
Seungeon Lee, Xiting Wang, Sungwon Han, Xiaoyuan Yi, Xing Xie, Meeyoung Cha
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 direct backbone prediction with SELOR logic-rule reasoning | BERT · Yelp | AUC: 97.39 -> 97.28 | No clear effect |
| Replace direct backbone prediction with SELOR logic-rule reasoning | RoBERTa · Clickbait | AUC: 63.72 -> 64.14 | Improved |
| Replace direct backbone prediction with SELOR logic-rule reasoning | DNN · Adult | AUC: 68.62 -> 70.36 | Improved |
| Use SELOR instead of RCN with a predefined rule set | Clickbait · Clickbait | AUC: 59.91 -> 60.31 (BERT); 59.37 -> 64.14 (RoBERTa) | Improved |
| Use SELOR logic-rule explanations instead of LIME explanations | Yelp · Yelp | good explanations (%): 89.8 -> 94.4 | Improved |
| Use SELOR logic-rule explanations instead of RCN explanations | Adult · Adult | best explanations (%): 10.2 -> 65.1 | Improved |
| Use the neural consequent estimator to assess rule confidence across the training data | SELOR · Noisy-label experiments | prediction performance: 2.4% to 13.7% | Improved |
| Exclude the candidate atom vegas during testing (ablation) | SELOR · 169 Yelp samples that previously used vegas | prediction performance | Mixed |
| Use predefined rules rather than SELOR's generated rules | RCN · Clickbait | AUC: 60.31 vs 59.91 (BERT); 64.14 vs 59.37 (RoBERTa) | Hurt |
Closest papers by meaning
- Learn to Explain Efficiently via Neural Logic Inductive Learning (ICLR 2020 · similarity 0.85)
- Towards Modeling Uncertainties of Self-Explaining Neural Networks via Conformal Prediction (AAAI 2024 · similarity 0.84)
- RULEBREAKERS: Challenging LLMs at the Crossroads between Formal Logic and Human-like Reasoning (ICML 2025 · similarity 0.84)
- Beyond Topological Self-Explainable GNNs: A Formal Explainability Perspective (ICML 2025 · similarity 0.83)
- Interpretable Concept-Based Memory Reasoning (NeurIPS 2024 · similarity 0.83)
- Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement (ICLR 2024 · similarity 0.83)
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.
