Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Efficient Equivariant Transfer Learning from Pretrained Models
Sourya Basu, Pulkit Katdare, Prasanna Sattigeri, Vijil Chenthamarakshan, Katherine Rose Driggs-Campbell, Payel Das, Lav R. Varshney
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 lambda-equitune instead of equitune (extra lambda network) | CLIP RN50 · CIFAR100 inference | inference time: 14.15 -> 23.5 | Hurt |
| use lambda-equitune instead of equitune (extra lambda network) | CLIP ViT-B/32 · CIFAR100 inference | inference time: 11.00 -> 16.08 | Hurt |
| use lambda-equitune instead of equitune | CLIP RN50 · CIFAR100 inference | memory: 3703 -> 3941 | Hurt |
| use equizero instead of equitune | CLIP RN50 · CIFAR100 inference | time and memory: 14.15 -> 14.10 | No clear effect |
| use equizero instead of equituning on non-equivariant pretrained LSTM | LSTM · SCAN Add Jump zero-shot | accuracy | Improved |
| use equizero instead of equituning on pretrained DQN | DQN · Gridworld, Cartpole, Acrobot | return | Improved |
| use equizero instead of equitune for GPT2 fairness | GPT2 · regard scores | regard | Improved |
| use equizero instead of equitune on CLIP with random rotations/flips | CLIP · ImageNet-V2 | accuracy | Improved |
| finetune with equizero vs equitune for 10k iterations | LSTM · SCAN | accuracy | Hurt |
| use equizero with max-probability or entropy loss on pretrained CNNs | Alexnet/Resnet · rot90-CIFAR10 | accuracy | Hurt |
| use lambda-equitune instead of equitune and equizero on CNNs | Alexnet/Resnet · rot90-CIFAR10 | accuracy | Improved |
| use lambda-equitune vs equizero on CLIP ViT-B/32 finetuning | CLIP ViT-B/32 · CIFAR100 rot90 | accuracy | Hurt |
Closest papers by meaning
- Improving Equivariant Model Training via Constraint Relaxation (NeurIPS 2024 · similarity 0.86)
- Approximately Equivariant Neural Processes (NeurIPS 2024 · similarity 0.85)
- Affine Equivariant Networks Based on Differential Invariants (CVPR 2024 · similarity 0.85)
- Learning (Approximately) Equivariant Networks via Constrained Optimization (NeurIPS 2025 · similarity 0.85)
- Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions (NeurIPS 2025 · similarity 0.85)
- Probing Equivariance and Symmetry Breaking in Convolutional Networks (NeurIPS 2025 · 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.
