AI Paper Atlas

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

Fingerprinting Pre-trained Encoders under Arbitrary Downstream Fine-Tuning via Adversarial Shifting

Tianlong Xu, Zixiong Wang, Lishuai Hou, Gaoyang Liu, Chen Wang, Xiaoyi Fan

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias

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
switch downstream adaptation from head-only (FTLL) to full fine-tuning (FTAL)Adversarial Shifting · SimCLR ResNet-50 -> STL-10Matching Rate: 0.960 -> 0.760Hurt
switch from FTLL to FTALSSL-WM · CIFAR-100Matching Rate: 0.484 -> 0.291Hurt
use label-only group fingerprint instead of label-bound fingerprintsAdversarial Shifting vs ADV-TRA/MFUE · FTLL, SimCLR STL-10Matching Rate: 0.426 (MFUE) -> 0.960Improved
use non-invasive fingerprint instead of watermark embeddingAdversarial Shifting vs SSL-WM · clean accuracyCDA: SSL-WM average loss 0.41%, ours noneImproved
use anchor-guided group fingerprintsAdversarial Shifting · CIFAR-100AUC: 0.84 (MEA) -> 0.96Improved
use surrogate-free generationAdversarial Shifting vs SSL-WM · CIFAR-100time: 681m -> 17mImproved
prune 60-70% of parametersAdversarial Shifting · GTSRB/CIFAR-100 (FTLL)Matching Rate: > 0.9No clear effect
prune above 0.8 ratioAdversarial Shifting · GTSRB/CIFAR-100Matching RateHurt
hard-label model extraction (DFMS-HL)Adversarial Shifting · GTSRB/CIFAR-100Matching Rate: < 0.3Hurt
embedding noise epsilon=0.1MEA · CIFAR-100Matching Rate: 0.43Hurt
embedding noiseAdversarial Shifting · CIFAR-100Matching Rate: > 0.8No clear effect
StolenEncoder extractionAdversarial Shifting · GTSRBMatching Rate: 0.6Mixed

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