Research map / Large language models / Agents, code and math
Improving Certified Robustness via Statistical Learning with Logical Reasoning
Zhuolin Yang, Zhikuan Zhao, Boxin Wang, Jiawei Zhang, Linyi Li, Hengzhi Pei, Bojan Karlaš, Ji Liu, Heng Guo, Ce Zhang, Bo 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 domain knowledge rules and MLN reasoning to randomized-smoothed sensing models | GTSRB road-sign sensing pipeline · GTSRB road signs, ℓ2 perturbations CI=0.12, 0.25, 0.50, and 1.00 | certified accuracy: 90.8 -> 96.0; 88.4 -> 91.0; 73.2 -> 75.5; 61.7 -> 62.5 | Improved |
| add domain knowledge rules and MLN reasoning | HighTech information-extraction pipeline · HighTech information extraction, CS=0.1 | certified accuracy: 99.7 -> 100.0 | Improved |
| add domain knowledge rules and MLN reasoning | HighTech information-extraction pipeline · HighTech information extraction, CS=0.5 | certified accuracy: 94.7 -> 100.0 | Improved |
| add domain knowledge rules and MLN reasoning | HighTech information-extraction pipeline · HighTech information extraction, CS=0.9 | certified accuracy: 38.4 -> 58.8 | Improved |
| add hierarchy and exclusion knowledge rules | PrimateNet sensing pipeline · PrimateNet benign test data, σ=0.12, 0.25, and 0.50 | accuracy: 0.9638 -> 0.9670; 0.9554 -> 0.9612; 0.9371 -> 0.9435 | Improved |
| add hierarchy and exclusion knowledge rules | PrimateNet sensing pipeline · PrimateNet, CI=0.12, α=10%, σ=0.12 | certified robustness: 0.5724 -> 0.8849 | Improved |
| add hierarchy and exclusion knowledge rules | PrimateNet sensing pipeline · PrimateNet, CI=0.50, α=50%, σ=0.50 | certified robustness: 0.4635 -> 0.5581 | Improved |
| use a knowledge-based sensing-reasoning pipeline rather than vanilla randomized smoothing | GTSRB road-sign classification · CI=0.50 or 1.00 with σ=0.12 | certified accuracy: 0.0 -> 73.2; 0.0 -> 24.2 | Improved |
Closest papers by meaning
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.86)
- Achieving Domain-Independent Certified Robustness via Knowledge Continuity (NeurIPS 2024 · similarity 0.85)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.85)
- COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuits (ICLR 2024 · similarity 0.85)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.85)
- Certified Robustness for Deep Equilibrium Models via Serialized Random Smoothing (NeurIPS 2024 · 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.
