AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Effective Robustness against Natural Distribution Shifts for Models with Different Training Data

Zhouxing Shi, Nicholas Carlini, Ananth Balashankar, Ludwig Schmidt, Cho-Jui Hsieh, Alex Beutel, Yao Qin

NeurIPS 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Zero-shot robustness and transfer · 0 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 two ID test sets (multi-ID) instead of oneeffective robustness fit · CIFAR-10.2R^2: 0.981 -> 0.996Improved
use two ID test sets instead of oneeffective robustness fit · CIFAR-10.2MAE (%): 2.22 -> 0.95Improved
use two ID test sets instead of oneImageNet models · CIFAR-10.2effective robustness (%): 3.91 -> 0.56Improved
use two ID test sets instead of onefit for LAION vs ImageNet · ImageNet-RR^2: 0.216 -> 0.982Improved
use two ID test sets instead of oneLAION models · ImageNet-like OOD averageeffective robustness (%): 5.96 -> -0.00Improved
use two ID test sets instead of oneYFCC models · ImageNet-like OOD averageeffective robustness (%): 2.59 -> 0.77Improved
switch the single ID test set from ImageNet to YFCCsingle-ID effective robustness · YFCC vs ImageNet modelseffective robustnessMixed
use two ID test sets instead of oneImageNet models (37) · ImageNet-Reffective robustness (%): -9.45 -> -0.54Improved
include mixed-data (ImageNet+YFCC) models in the baseline fit (ablation)multi-ID fit · ObjectNetMAE (%): 1.73 -> 0.95Improved
include mixed-data (ImageNet+LAION) models in the baseline fit (ablation)multi-ID fit · ImageNet-SketchMAE (%): 2.62 -> 2.03Improved
use two ID test sets instead of oneOpenCLIP LAION · ImageNet-like OOD averageMAE (%): 22.59 -> 2.63Improved
use two ID test sets instead of oneSLIP YFCC · ImageNet-like OOD averageeffective robustness (%): 10.56 -> 3.19Mixed

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