Research map / Architectures and efficiency / RNNs and normalization
Knowledge-Adaptation Priors
Mohammad Emtiyaz Khan, Siddharth Swaroop
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 K-priors instead of Replay with the same memory | K-priors · UCI Adult, USPS odd vs even, USPS neural networks | classification performance | Improved |
| Use K-priors instead of full-batch retraining with limited memory | K-priors · UCI Adult and USPS odd vs even adaptation tasks | classification performance | No clear effect |
| Use K-priors instead of full-batch retraining with limited memory | K-priors · CIFAR-10 Add Data | classification performance | Hurt |
| Add a temperature to K-priors (ablation) | K-priors · CIFAR-10 Add Data | classification performance | Improved |
| Add a temperature to K-priors (ablation) | K-priors · CIFAR-10 Change Architecture | classification performance | Improved |
| Add the weight-space term centered at the base model (ablation) | K-priors · USPS logistic regression adaptation tasks | classification performance | Improved |
| Replace the centered weight term with an uncentered weight penalty (ablation) | K-priors · USPS logistic regression adaptation tasks | classification performance | Hurt |
| Randomly initialize parameters instead of initializing at the previous parameters (ablation) | K-priors · USPS odd vs even neural networks: Add Data, Remove Data, Change Regularizer | classification performance | No clear effect |
| Choose Replay memory randomly instead of by highest h'(base prediction) (ablation) | Replay · UCI Adult Add Data | classification performance | No clear effect |
| Use weight-priors instead of K-priors (ablation) | Add Data adaptation · USPS odd vs even, logistic regression and neural networks | classification performance | Hurt |
| Use weight-priors instead of K-priors (ablation) | Add Data adaptation · UCI Adult, MNIST, and CIFAR-10 | classification performance | No clear effect |
Closest papers by meaning
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.85)
- C-LoRA: Continual Low-Rank Adaptation for Pre-trained Visual Models (NeurIPS 2026 · similarity 0.85)
- Continual Knowledge Adaptation for Reinforcement Learning (NeurIPS 2025 · similarity 0.85)
- PaLoRA: Paced Low-Rank Adaptation for Continual Learning (NeurIPS 2026 · similarity 0.85)
- A Bayesian Model Selection Criterion for Selecting Pretraining Checkpoints (ICML 2025 · similarity 0.85)
- Solving New Tasks by Adapting Internet Video Knowledge (ICLR 2025 · similarity 0.85)
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.
