Research map / Architectures and efficiency / Attention and transformers
Dynamic Inference with Neural Interpreters
Nasim Rahaman, Muhammad Waleed Gondal, Shruti Joshi, Peter Vincent Gehler, Yoshua Bengio, Francesco Locatello, Bernhard Schölkopf
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 trainable type-matching parameters to CLS-token finetuning (ablation) | Neural Interpreter · Fuzzy Boolean adaptation | R2: 0.9202 -> 0.9857 | Improved |
| also finetune remaining parameters beyond type matching (ablation) | Neural Interpreter · Fuzzy Boolean adaptation | R2: 0.9857 -> 0.9953 | Improved |
| use Neural Interpreter instead of ViT | PGM model · PGM Neutral test | accuracy: 72.7 -> 77.0 | Improved |
| use Neural Interpreter instead of ViT | PGM model · PGM Triples test | accuracy: 15.9 -> 20.0 | Improved |
| use Neural Interpreter instead of ViT | PGM model · PGM Interpolation val | accuracy: 89.9 -> 87.9 | Hurt |
| keep pretrained functions instead of randomizing codes and signatures before adaptation (ablation) | Neural Interpreter · K-MNIST adaptation | accuracy | Improved |
| add more functions at adaptation time | Neural Interpreter · K-MNIST adaptation | accuracy | Improved |
| use Neural Interpreter instead of ViT for low-data adaptation | digits pretrained model · K-MNIST fast adaptation | accuracy | Improved |
Closest papers by meaning
- Attention over Learned Object Embeddings Enables Complex Visual Reasoning (NeurIPS 2021 · similarity 0.88)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.86)
- NePTune: A Neuro-Pythonic Framework for Tunable Compositional Reasoning on Vision-Language (ICLR 2026 · similarity 0.86)
- Neurons to Words: A Novel Method for Automated Neural Network Interpretability and Alignment (AAAI 2025 · similarity 0.85)
- Towards flexible perception with visual memory (ICML 2025 · similarity 0.84)
- A Multimodal Automated Interpretability Agent (ICML 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.
