A test-time adaptation framework anchors adversarial training to a non-robust teacher's predictions, yielding more stable optimization and better robustness-accuracy trade-offs than standard self-consistency methods.
Theoretically principled trade-off between robustness and accuracy
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
verdicts
UNVERDICTED 4representative citing papers
Attention Hijacking is a new attack that improves cross-query transferability in VLMs by explicitly steering internal attention to a persistent image-dominant pattern.
Negative-capable ridge regression uses controlled negative regularization as anti-shrinkage to increase effective complexity along weak eigendirections and mitigate underfitting in small-data regression.
AGC is a training-free inference-time defense for CLIP that adaptively corrects features along geodesics to robust augmentations, claiming 44.4% higher average robust accuracy and 10x lower latency than prior baselines across eight datasets and three backbones.
citing papers explorer
-
Learning Robustness at Test-Time from a Non-Robust Teacher
A test-time adaptation framework anchors adversarial training to a non-robust teacher's predictions, yielding more stable optimization and better robustness-accuracy trade-offs than standard self-consistency methods.
-
Attention Hijacking: Response Manipulation Across Queries in Vision-Language Models
Attention Hijacking is a new attack that improves cross-query transferability in VLMs by explicitly steering internal attention to a persistent image-dominant pattern.
-
A Ridge Too Far: Correcting Over-Shrinkage via Negative Regularization
Negative-capable ridge regression uses controlled negative regularization as anti-shrinkage to increase effective complexity along weak eigendirections and mitigate underfitting in small-data regression.
-
AGC: Adaptive Geodesic Correction for Adversarial Robustness on Vision-Language Models
AGC is a training-free inference-time defense for CLIP that adaptively corrects features along geodesics to robust augmentations, claiming 44.4% higher average robust accuracy and 10x lower latency than prior baselines across eight datasets and three backbones.