↓Skip to main content

🏢 Institute of Computing Technology, Chinese Academy of Sciences

UMFC: Unsupervised Multi-Domain Feature Calibration for Vision-Language Models
·2814 words·14 mins· loading · loading
AI Generated Multimodal Learning Vision-Language Models 🏢 Institute of Computing Technology, Chinese Academy of Sciences
UMFC: Unsupervised Multi-domain Feature Calibration improves vision-language model transferability by mitigating inherent model biases via a novel, training-free feature calibration method.
TopoLogic: An Interpretable Pipeline for Lane Topology Reasoning on Driving Scenes
·2304 words·11 mins· loading · loading
AI Generated AI Applications Autonomous Vehicles 🏢 Institute of Computing Technology, Chinese Academy of Sciences
TopoLogic uses lane geometry and query similarity to improve lane topology reasoning in autonomous driving, significantly outperforming existing methods.
Self-Supervised Adversarial Training via Diverse Augmented Queries and Self-Supervised Double Perturbation
·2025 words·10 mins· loading · loading
Machine Learning Self-Supervised Learning 🏢 Institute of Computing Technology, Chinese Academy of Sciences
DAQ-SDP enhances self-supervised adversarial training by using diverse augmented queries, a self-supervised double perturbation scheme, and a novel Aug-Adv Pairwise-BatchNorm method, bridging the gap …
Leveraging Catastrophic Forgetting to Develop Safe Diffusion Models against Malicious Finetuning
·2397 words·12 mins· loading · loading
🏢 Institute of Computing Technology, Chinese Academy of Sciences
This paper introduces a novel training policy that leverages catastrophic forgetting to make diffusion models resilient against malicious fine-tuning, effectively preventing the generation of harmful …
Continual Learning in the Frequency Domain
·2169 words·11 mins· loading · loading
Machine Learning Continual Learning 🏢 Institute of Computing Technology, Chinese Academy of Sciences
Boost continual learning efficiency with CLFD: a novel frequency domain approach that improves accuracy by up to 6.83% and slashes training time by 2.6x on edge devices!