Attention-Sink-Aware Hallucination Attacks and Defenses on Multimodal LLMs Spring 2026 · POSTECH · AI Security coursePurple TeamingMultimodal LLM
Optimization Methods for Improving the Accuracy of Homomorphically Encrypted LLM Inference 2026 · LG ElectronicsHomomorphic EncryptionLLM
Optimizing Homomorphic Computations for Transformer Models 2025 · LG ElectronicsHomomorphic EncryptionTransformer
Building and Optimizing a GPU-based Homomorphic Encryption Library 2023 · Samsung Advanced Institute of TechnologyHomomorphic EncryptionGPU Acceleration