New TIMA framework enhances zero-shot adversarial robustness in foundation models

By PulseAugur Editorial · [1 sources] · 2026-08-18 04:00

Researchers have developed a new framework called TIMA (Text-Image Mutual Awareness) to improve the zero-shot adversarial robustness of foundation models like CLIP. TIMA addresses challenges in maintaining generalization ability while enhancing robustness against adversarial attacks. The framework includes modules for tuning text and image embeddings to better balance logit margins and preserve semantic consistency, outperforming existing methods in experiments. AI

IMPACT This research could lead to more resilient AI models capable of handling adversarial attacks without compromising their general performance.

RANK_REASON The cluster describes a research paper detailing a new framework for improving AI model robustness. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

New TIMA framework enhances zero-shot adversarial robustness in foundation models

COVERAGE [1]

arXiv cs.AI TIER_1 English(EN) · Fengji Ma, Hei Victor Cheng, Chenxing Li, Li Liu · 2026-08-18 04:00

TIMA: Text-Image Mutual Awareness for Balancing Zero-Shot Adversarial Robustness and Generalization Ability

arXiv:2405.17678v2 Announce Type: replace-cross Abstract: Achieving zero-shot adversarial robustness without sacrificing generalization remains challenging for foundation models such as CLIP, especially under large adversarial perturbations. Through empirical analyses, we identif…

COVERAGE [1]

TIMA: Text-Image Mutual Awareness for Balancing Zero-Shot Adversarial Robustness and Generalization Ability

RELATED ENTITIES

RELATED TOPICS