Research & Innovation

Global: New Framework Enables Continual Unlearning for Large Language Models While Preserving UtilityBackground and MotivationResearchers have introduced a framework called fit that tackles the growing demand for systematic removal of…

Global: Quantum Continual Learning Framework for Intrusion DetectionA team of researchers has introduced QCL-IDS, a quantum‑centric continual‑learning system designed to detect network intrusions while meeting strict operational limits on compute,…

Global: Study Finds Compression-Aware Attacks Reduce Robustness of Vision-Language Models A team of artificial‑intelligence researchers announced on January 2026 that a newly proposed adversarial method, called Compression‑AliGnEd (CAGE), significantly lowers…

Global: SecIC3 Boosts Hardware Non-Interference Verification SpeedA new hardware model‑checking algorithm named SecIC3 has been introduced to accelerate the verification of confidentiality and integrity properties in digital designs. The approach…

Global: Researchers Connect Sensitivity Awareness in Quantized LLMs to Differential PrivacyA team of AI researchers announced on arXiv that they have formalized “sensitivity awareness” for large language models (LLMs) and…

Global: LLM Routing Systems Found Vulnerable to Adversarial Rerouting; New Guardrail Framework Offers 99% Detection AccuracyBackground on Multi-Model AI RoutingRecent research highlights the growing use of large language model (LLM)…

Global: Study Demonstrates Single Perturbation Can Hijack Multimodal LLM DecisionsResearchers from five institutions reported that a single, universal adversarial perturbation can manipulate the outputs of multimodal large language models (MLLMs)…

Global: HE‑Optimized Neural Network Architecture Delivers Up to 9.78× Faster InferenceIn January 2026, researchers announced StriaNet, a neural‑network design specifically engineered for homomorphic encryption (HE) workloads, achieving speedups of 9.78×…

Global: Researchers Identify High-Frequency Bias in Membership Inference Attacks on Diffusion ModelsScientists have uncovered a systematic bias that affects the accuracy of membership inference attacks (MIAs) targeting diffusion‑based image generators.…

Global: TrajPrint Enables Lossless Black-Box Verification of Diffusion Model CopyrightResearchers have introduced a new framework called TrajPrint that allows lossless verification of diffusion model copyright in black‑box API settings. The…