Research map / Architectures and efficiency / RNNs and normalization
Scaling Symbolic Methods using Gradients for Neural Model Explanation
Subham Sekhar Sahoo, Subhashini Venugopalan, Li Li, Rishabh Singh, Patrick Riley
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 |
|---|---|---|---|
| Use SMT minimization to sparsify the IG-selected input mask (ablation) | SMUG · ImageNet validation | Sparsity%: 43.3 ± 0.32 -> 17.7 ± 0.10 | Improved |
| Use SMT minimization rather than the unminimized top-neuron receptive-field mask (ablation) | SMUG · ImageNet validation | LSC: -1.23 -> -1.26 | Improved |
| Use SMUG instead of pixel-space Integrated Gradients | Inception-v1 explanations · ImageNet validation | LSC: -0.29 -> -1.26 | Improved |
| Increase the number of selected top-k neurons (ablation) | SMUG · ImageNet | mask size | Hurt |
| Increase the number of selected top-k neurons (ablation) | SMUG · ImageNet | solver runtime | Hurt |
| Decrease gamma (ablation) | SMUG · ImageNet | mask size | Improved |
| Decrease gamma (ablation) | SMUG · ImageNet | solver runtime | Hurt |
| Encode the full network in SMT rather than selecting first-layer neurons with gradient information | SMT-based explanation on MNIST · MNIST validation | solved instances: 34 / 100 | Hurt |
Closest papers by meaning
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.86)
- Transcoders find interpretable LLM feature circuits (NeurIPS 2024 · similarity 0.85)
- Alternating Gradient Flows: A Theory of Feature Learning in Two-layer Neural Networks (NeurIPS 2025 · similarity 0.85)
- 4+3 Phases of Compute-Optimal Neural Scaling Laws (NeurIPS 2024 · similarity 0.85)
- Neurons to Words: A Novel Method for Automated Neural Network Interpretability and Alignment (AAAI 2025 · similarity 0.84)
- Scaling Laws for Sparsely-Connected Foundation Models (ICLR 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.
