AI Paper Atlas

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

NeurIPS 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Zero-shot robustness and transfer · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
use lambda-equitune instead of equitune (extra lambda network)CLIP RN50 · CIFAR100 inferenceinference time: 14.15 -> 23.5Hurt
use lambda-equitune instead of equitune (extra lambda network)CLIP ViT-B/32 · CIFAR100 inferenceinference time: 11.00 -> 16.08Hurt
use lambda-equitune instead of equituneCLIP RN50 · CIFAR100 inferencememory: 3703 -> 3941Hurt
use equizero instead of equituneCLIP RN50 · CIFAR100 inferencetime and memory: 14.15 -> 14.10No clear effect
use equizero instead of equituning on non-equivariant pretrained LSTMLSTM · SCAN Add Jump zero-shotaccuracyImproved
use equizero instead of equituning on pretrained DQNDQN · Gridworld, Cartpole, AcrobotreturnImproved
use equizero instead of equitune for GPT2 fairnessGPT2 · regard scoresregardImproved
use equizero instead of equitune on CLIP with random rotations/flipsCLIP · ImageNet-V2accuracyImproved
finetune with equizero vs equitune for 10k iterationsLSTM · SCANaccuracyHurt
use equizero with max-probability or entropy loss on pretrained CNNsAlexnet/Resnet · rot90-CIFAR10accuracyHurt
use lambda-equitune instead of equitune and equizero on CNNsAlexnet/Resnet · rot90-CIFAR10accuracyImproved
use lambda-equitune vs equizero on CLIP ViT-B/32 finetuningCLIP ViT-B/32 · CIFAR100 rot90accuracyHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy