Research map / Architectures and efficiency / RNNs and normalization
ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior
Florian Eichin, Yupei Du, Philipp Mondorf, Maria Matveev, Barbara Plank, Michael A. Hedderich
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 100 instead of 10 integration steps in the test feature map (ablation) | EPK representation · MOD-113 Transformer | EPK accuracy: 0.748 -> 1.0 | Improved |
| use 100 instead of 10 integration steps (ablation) | EPK representation · MOD-113 Transformer | KL divergence: 0.885 -> 0.0 | Improved |
| use 100 instead of 10 integration steps (ablation) | EPK representation · CIFAR-2 ResNet9 | EPK accuracy: 0.997 -> 1.0 | Improved |
| use 100 instead of 10 integration steps (ablation) | EPK representation · MNIST CNN | EPK accuracy: 0.998 -> 1.0 | Improved |
| prune CNN weights by ExPLAIND kernel importance (training-free) vs retrain-and-prune baseline | CNN, MNIST · MNIST CNN pruning 70% to 99% sparsity | test accuracy | No clear effect |
| prune CNN weights by ExPLAIND kernel importance vs magnitude-pruning baseline | CNN, MNIST · MNIST CNN pruning | KL divergence to original model | Improved |
| keep only top-ranked update steps (about 60% sparsity) in decomposition (ablation) | ExPLAIND · MNIST CNN | reconstructed prediction accuracy | No clear effect |
| initialize Attention and Linear-1 layers from grokked model, rest random (ablation) | Grokking Transformer · MOD-113 | generalization speed | Improved |
| start from earlier-checkpoint representation pipeline instead of final (ablation) | Grokking Transformer · MOD-113 | final accuracy | Hurt |
| swap grokked outer layers into intermediate checkpoints (ablation) | Grokking Transformer · MOD-113 | predictions of checkpoints | Improved |
| use data influence scores derived from ExPLAIND for causal data attribution | ExPLAIND · CIFAR and MNIST | data attribution | No clear effect |
Closest papers by meaning
- Local vs. Global Interpretability: A Computational Complexity Perspective (ICML 2024 · similarity 0.87)
- Additive Models Explained: A Computational Complexity Approach (NeurIPS 2025 · similarity 0.87)
- How to Probe: Simple Yet Effective Techniques for Improving Post-hoc Explanations (ICLR 2025 · similarity 0.87)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.86)
- Data-centric Prediction Explanation via Kernelized Stein Discrepancy (ICLR 2025 · similarity 0.86)
- Unraveling the Influence of Training Data and Internal Structures in Large Language Models for Enhanced Explainability (Student Abstract) (AAAI 2025 · similarity 0.86)
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.
