deepfake detection

South Korea's Forensic Service Achieves Rapid Deepfake Detection

South Korea's National Forensic Service has developed technology capable of detecting AI-generated deepfakes in seconds, while authorities warn against public release of the detection methods.

LLM safety

Q-Realign: New Method Restores LLM Safety During Quantization

Researchers introduce Q-realign, a technique that piggybacks safety realignment onto quantization, solving the problem of safety degradation in compressed LLMs for efficient deployment.

LLM Architecture

MoE-LoRA Framework Advances Multi-Task LLM Specialization

New research combines Mixture-of-Experts with Low-Rank Adaptation to create specialized AI models that maintain generalist capabilities while excelling at domain-specific tasks.

Deep Learning

NOVAK: A Unified Adaptive Optimizer for Deep Neural Networks

New research introduces NOVAK, a unified framework that bridges popular adaptive optimizers like Adam and AdaGrad, potentially improving training efficiency for deep learning models.

LLM compression

Hierarchical Sparse Plus Low Rank: A New Approach to LLM Compress

New research introduces hierarchical sparse plus low rank compression for LLMs, combining structured sparsity with matrix decomposition for efficient model deployment.

LLM Security

STAR Method Detects Hidden Backdoors in LLM Reasoning Chains

New research introduces State-Transition Amplification Ratio (STAR) to identify inference-time backdoor attacks in large language models by analyzing anomalous reasoning patterns.

Synthetic Media

Synthetic Speech Dialogues Enable New Mental Manipulation Detecti

New research uses LLM-generated multi-speaker dialogues to train AI systems that detect psychological manipulation in speech, advancing synthetic media analysis and content authenticity verification.

LLM safety

Logic-Guided Synthesis Tackles LLM Regulatory Compliance

New research introduces a framework for evaluating implicit regulatory compliance in LLM tool invocations using logic-guided synthesis, addressing critical AI safety concerns.

AI Detection

Linguistic Analysis Reveals Why AI Text Detectors Generalize

New research uncovers the linguistic patterns that enable AI-generated text detectors to generalize across different language models, offering insights for more robust synthetic content detection.

Kling AI

Kling AI Hits $240M Annual Revenue Run Rate in December 2025

Chinese AI video generator Kling AI reaches $240M annualized revenue run rate, signaling strong commercial demand for synthetic video tools.

LLM

Universal Latent Space Enables Zero-Shot LLM Routing

New research introduces a universal latent space approach for cost-efficient LLM routing, enabling zero-shot model selection without task-specific training data or expensive benchmarking.

LLM Infrastructure

AIConfigurator Speeds Up LLM Serving Optimization Dramatically

New research introduces AIConfigurator, a system that dramatically accelerates configuration optimization for multi-framework LLM serving, enabling faster deployment of AI inference infrastructure.