remove hyeonsu's paper from publication

Author: Yurim990507

_data/Publications_data.json

@@ -148,24 +148,6 @@
         "img": "/Publications/2025_CIKM_sangjun.jpg",
         "abstract": "Machine unlearning has emerged as a significant research topic in response to the increasing demands for data privacy and compliance with privacy regulations. The main challenge is to eliminate the influence of a specific subset of training data from a pretrained model while preserving the model’s performance on the retain set without retraining it from scratch. In this paper, we propose a novel efficient unlearning framework based on Optimal Transport, which can effectively work on class and instance-wise unlearning tasks. By analyzing and comparing the feature spaces of the original and retrained models, we formulate the unlearning problem as a distribution alignment task between the forget set and the retain set. We guide the feature distribution of the forget set, which initially forms distinct, structured patterns, to align with that of the retain set. In addition, we introduce a class-aware cost function for optimal transport that encourages inter-class transport, thereby enhancing the forgetting process. Extensive experiments on three public benchmark datasets demonstrate its superior effectiveness compared to previous SOTA methods."
     },
-    {
-        "title": "FASE: Feature-Aligned Scene Encoding for Open-Vocabulary Object Detection in Remote Sensing",
-        "authors": [
-            "Hyeonsu Hwang",
-            "Simon S. Woo"
-        ],
-        "venue_full": "ACM International Conference on Information and Knowledge Management",
-        "venue": "CIKM",
-        "track": "Short",
-        "Factor": [
-            "BK Computer Science IF=",
-            2
-        ],
-        "year": 2025,
-        "links": {},
-        "img": "/Publications/2025_CIKM_Hyeonsu.png",
-        "abstract": "Open-vocabulary object detection (OVD) in remote sensing (RS) has shown remarkable generalization capabilities across diverse RS imagery through alignment between image and text embeddings. Such methods have further improved detection performance by incorporating additional scene-level context from both visual and textual domains. However, existing methods approximate scene context by simply averaging the text embeddings of the image’s object labels, which is insufficient to capture the rich linguistic context present in RS scenes. To address this limitation, we propose a novel Feature-Aligned Scene Encoding (FASE), which constructs comprehensive scene representations through high-quality captions generated by a specialized vision-language model. These caption embeddings are then aligned with visual features through a Feature Alignment Module (FAM) that employs dual-branch fusion with gating and cross-attention mechanisms. By utilizing enhanced scene encoding only during training, our method internalizes rich contextual knowledge while maintaining inference efficiency. Experiments on multiple benchmarks demonstrate significant improvements over state-of-the-art methods, validating the effectiveness of our approach for OVD in RS."
-    },
     {
         "title": "Beyond Masking: Landmark-based Representation Learning and Knowledge-Distillation for Audio-Visual Deepfake Detection",
         "authors": [