Jingfeng Zhang

Researcher@RIKEN·日本·OpenReview

研究方向

Multimodal Models · Trustworthy Embodied AI and Robotics · Trustworthy machine learning · Adversarial robustness · Deep learning in the view of the dynamical systems

Jingfeng Zhang

ALMEA: Active Learning-Enhanced Multimodal Entity Alignment with Semantic Modality Imputation

Dissecting Representation Misalignment in Contrastive Learning via Influence Function

Unlearning during Training: Domain-Specific Gradient Ascent for Domain Generalization

Benign Overfitting in Adversarial Training for Vision Transformers

Robust Learning of Diffusion Models with Extremely Noisy Conditions

Towards Robust Agentic Systems through Generative Flow Exploration of Primitives

Concept-Based Dictionary Learning for Inference-Time Safety in Vision–Language–Action Models

Multi-agent Imitates Enterprise Dynamics

Goal-oriented Backdoor Attack against Vision-Language-Action Models via Physical Objects

Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models

Editable Concept Bottleneck Models

Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence

DDAD: A Two-Pronged Adversarial Defense Based on Distributional Discrepancy

One Stone, Two Birds: Enhancing Adversarial Defense Through the Lens of Distributional Discrepancy

Editable Concept Bottleneck Models

Learning without Isolation: Pathway Protection for Continual Learning

Dissecting Misalignment of Multimodal Large Language Models via Influence Function

Balancing Domain-Invariant and Domain-Specific Knowledge for Domain Generalization with Online Knowledge Distillation

Do We Really Need Parameter-Isolation to Protect Task Knowledge?