Pulling Back the Curtain on Deep Networks
cs.LG, cs.CV, cs.NE
Submitted: 2025-07-30
Updated: 2026-11-02
Code: https://github.com/314-Foundation/SemanticPullbacks
Terminology
Sources
- Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
- SmoothGrad: removing noise by adding noise
- "Why Should I Trust You?": Explaining the Predictions of Any Classifier
- Axiomatic Attribution for Deep Networks
- A Unified Approach to Interpreting Model Predictions
- Learning Important Features Through Propagating Activation Differences
- Interpretation of Neural Networks is Fragile
- Sanity Checks for Saliency Maps
- How Well do Feature Visualizations Support Causal Understanding of CNN Activations?
- Don't trust your eyes: on the (un)reliability of feature visualizations
- Attribution Explanations for Deep Neural Networks: A Theoretical Perspective
- Fooling Neural Network Interpretations via Adversarial Model Manipulation
- Explanations can be manipulated and geometry is to blame
- B-cosification: Transforming Deep Neural Networks to be Inherently Interpretable
- Deep Residual Learning for Image Recognition
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions
- Captum: A unified and generic model interpretability library for PyTorch
- On the (In)fidelity and Sensitivity for Explanations
- Towards Robust Interpretability with Self-Explaining Neural Networks
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis