AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Topology-Aware Contrastive Learning: Regulating Representation Connectivity via Persistent Homology

Jiaxin Sun, Yuhua Qian, Yang Wang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory

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
add topology-aware connectivity regulationSimCLR · ImageNet-100k-NN purity (k=3): 30.59 -> 37.19Improved
add topology-aware connectivity regulationSimCLR · ImageNet-100linear accuracy: 60.98 -> 64.14Improved
add topology-aware connectivity regulationSimCLR · CIFAR-100k-NN purity (k=3): 57.73 -> 60.36Improved
add topology-aware connectivity regulationSimCLR · CIFAR-100linear accuracy: 60.38 -> 62.47Improved
add topology-aware connectivity regulationSimCLR · CIFAR-10k-NN purity (k=3): 71.24 -> 71.77Improved
add topology-aware connectivity regulationSimCLR · STL-10k-NN purity (k=3): 74.59 -> 75.27Improved
add topology-aware connectivity regulation under weak augmentationSimCLR · CIFAR-100 weak (a)+(c)Top-1 accuracy: 38.92 -> 43.19Improved
add topology-aware connectivity regulation under strong augmentationSimCLR · CIFAR-100 strong (b)+(d)Top-1 accuracy: 53.70 -> 55.72Improved
add topology-aware connectivity regulation under strong color jitterSimCLR · CIFAR-100 strong (d)Top-1 accuracy: 11.33 -> 17.35Improved
add topology-aware connectivity regulation under weak cropSimCLR · CIFAR-100 weak (a)Top-1 accuracy: 32.98 -> 35.13Improved
add topology-aware connectivity regulation on Vehicles 1 superclass with k=20SimCLR · CIFAR-100 Vehicles 1k-NN purity (k=20): 59.07 -> 60.49Improved
add topology-aware connectivity regulation on Fish superclass with k=8SimCLR · CIFAR-100 Fishk-NN purity (k=8): 61.25 -> 62.40Improved

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