Advances in Robust AI: Defending Adversarial Attacks, Boosting Domain Generalization, Stopping LLM Jailbreaks

This article reviews the latest progress in designing algorithms with strong robustness, covering adversarial examples in computer vision, novel training paradigms and certification methods, domain‑generalization techniques that achieve state‑of‑the‑art performance in medical imaging and molecular recognition, and new attack‑defense strategies for LLM jailbreak scenarios.

AI SafetyLLM Securityadversarial robustness

0 likes · 4 min read

Advances in Robust AI: Defending Adversarial Attacks, Boosting Domain Generalization, Stopping LLM Jailbreaks

AI Frontier Lectures

Sep 8, 2025 · Artificial Intelligence

Why Data Augmentation Triggers OOD Fluctuations and How PEER Solves It

Data augmentation, while popular for single-source domain generalization, often induces severe out-of-distribution performance swings during training; the PEER framework combats this by employing dual-model collaboration, entropy regularization, periodic parameter averaging, and dynamic augmentation, achieving state-of-the-art robustness across multiple benchmark datasets.

OOD robustnessdata augmentationdomain generalization

0 likes · 7 min read

Why Data Augmentation Triggers OOD Fluctuations and How PEER Solves It

Kuaishou Tech

Apr 18, 2022 · Artificial Intelligence

SSAN: A Novel Dual‑Stream Network for Domain‑Generalized Face Anti‑Spoofing

This paper proposes SSAN, a novel dual‑stream network that separates content and style features to achieve domain‑generalized face anti‑spoofing, employing adversarial learning for content, contrastive learning for style, and a large‑scale evaluation protocol across twelve public datasets, achieving state‑of‑the‑art performance.

SSANStyle Transfercontrastive learning

0 likes · 16 min read

SSAN: A Novel Dual‑Stream Network for Domain‑Generalized Face Anti‑Spoofing

JD Tech Talk

Oct 12, 2020 · Artificial Intelligence

Transfer Learning for Human Mobility Modeling in New Cities

The paper presented at WWW 2020 proposes a transfer‑learning framework that leverages POI, road‑network and traffic data from existing cities to generate realistic human mobility trajectories for a target city by modeling mobility intentions, origin‑destination pairs, and routes, and validates the approach with extensive experiments across multiple Chinese cities.

Urban Computingaidomain generalization

0 likes · 10 min read

Transfer Learning for Human Mobility Modeling in New Cities