AEGIS localizes sparse semantic-injecting attention heads in diffusion models and applies similarity-aware repulsion at those heads to block visual synonym jailbreaks while preserving benign generation.
Microsoft coco: Common objects in context
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
representative citing papers
ELSA is a near-SRAM dataflow architecture realizing elastic inference in SNNs via fine-grained spine/token pipelines, bundled AER, and mini-batch Gustavson products, delivering up to 3.4x speedup and 22.1x energy gains over SOTA accelerators on ResNet-50.
Hydra stabilizes multi-concept backdoor attacks in diffusion models via evolutionary trigger search in text encoder space and trigger-clean regularization during multi-task fine-tuning, achieving high attack success while preserving clean image quality.
SGP-Net improves few-shot medical image segmentation by using spectral frequency decomposition for cue disentanglement and geodesic matching on feature manifolds instead of cosine similarity.
RGSE adapts text embeddings at test time via evolutionary search, using cosine similarity rewards from high-confidence visual proposals to improve open-vocabulary object detection under distribution shifts.
NAN-SPOT detects unknown objects better than retraining-heavy methods by using Negative-Aware Norm from off-the-shelf detectors and introduces the expanded COCO-Open dataset.
A modular system fuses object detection, segmentation, and LiDAR-improved depth estimation to achieve 0.63 m MAE for obstacle distances on synthetic railway data.
DeepDetect trains ESPNet on fused classical detector masks to produce dense, repeatable keypoints that outperform prior methods on Oxford, HPatches, and Middlebury benchmarks.
citing papers explorer
-
AEGIS: A Mechanism-Guided Defense against Visual Synonym Jailbreaks in Text-to-Image Models
AEGIS localizes sparse semantic-injecting attention heads in diffusion models and applies similarity-aware repulsion at those heads to block visual synonym jailbreaks while preserving benign generation.
-
ELSA: An ELastic SNN Inference Architecture for Efficient Neuromorphic Computing
ELSA is a near-SRAM dataflow architecture realizing elastic inference in SNNs via fine-grained spine/token pipelines, bundled AER, and mini-batch Gustavson products, delivering up to 3.4x speedup and 22.1x energy gains over SOTA accelerators on ResNet-50.
-
Awakening the Hydra: Stabilizing Multi-Concept Backdoor Injection in Text-to-Image Diffusion Models
Hydra stabilizes multi-concept backdoor attacks in diffusion models via evolutionary trigger search in text encoder space and trigger-clean regularization during multi-task fine-tuning, achieving high attack success while preserving clean image quality.
-
Beyond Euclidean Prototypes: Spectral Disentanglement and Geodesic Matching for Few-Shot Medical Image Segmentation
SGP-Net improves few-shot medical image segmentation by using spectral frequency decomposition for cue disentanglement and geodesic matching on feature manifolds instead of cosine similarity.
-
Reward-Guided Semantic Evolution for Test-time Adaptive Object Detection
RGSE adapts text embeddings at test time via evolutionary search, using cosine similarity rewards from high-confidence visual proposals to improve open-vocabulary object detection under distribution shifts.
-
Beyond Known Objects: A Novel Framework for Open-Set Object Detection using Negative-Aware Norm
NAN-SPOT detects unknown objects better than retraining-heavy methods by using Negative-Aware Norm from off-the-shelf detectors and introduces the expanded COCO-Open dataset.
-
Integrating Object Detection, LiDAR-Enhanced Depth Estimation, and Segmentation Models for Railway Environments
A modular system fuses object detection, segmentation, and LiDAR-improved depth estimation to achieve 0.63 m MAE for obstacle distances on synthetic railway data.
-
DeepDetect: Learning All-in-One Dense Keypoints
DeepDetect trains ESPNet on fused classical detector masks to produce dense, repeatable keypoints that outperform prior methods on Oxford, HPatches, and Middlebury benchmarks.
- Progressive Semantic Communication for Efficient Edge-Cloud Vision-Language Models