Research & Innovation

Global: Kernel-Level DVFS Achieves Up to 14.6% Energy Savings in GPT-3 TrainingThe authors of a recent arXiv preprint have demonstrated that a fine‑grained, kernel‑level dynamic voltage and frequency scaling (DVFS)…

Global: Two-Stage Transformer Model Improves Functional Group Removal and ReplacementA team of chemoinformatics researchers announced a novel two‑stage transformer architecture designed to remove and replace functional groups in chemical compounds.…

Global: FCNN-Driven Non-Local Electron Pressure Closure Enhances Kinetic Plasma SimulationsA team of plasma physicists has introduced a non‑local five‑moment electron pressure tensor closure that leverages a Fully Convolutional Neural Network…

Global: Vector-Injected In-Context Learning Enhances Forecasting with Frozen Large Language ModelsResearchers have introduced a new method called LVICL that improves time series forecasting performance while keeping all parameters of a…

Global: Mamba-Based RNNs Exhibit Limited Forgetting Capability in Long-Context ScenariosResearchers who authored a paper posted on arXiv in 2024 report that Mamba-style recurrent neural networks (RNNs) fail to effectively discard…

Global: Study Introduces Decompositional Training to Enhance LLM Algorithmic Capabilities Researchers at the SyCoSMA laboratory within LIRIS announced on Jan. 12, 2026 that they have developed a supervised training framework,…

Global: Study Reveals Dynamic Attention Patterns in Masked Diffusion ModelsA team of researchers led by Xin Dai and colleagues published a paper on January 12, 2026 that examines how attention…

Global: Subgraph-Informed Scoring Boosts Knowledge Graph Completion AccuracyOverviewA new framework named SLogic has been introduced to assign query‑dependent scores to logical rules used in knowledge graph completion. By evaluating the…

Global: Sherry Framework Enhances Ternary Quantization for Edge Deployment of Large Language ModelsResearchers have introduced Sherry, a hardware‑efficient ternary quantization framework designed to lower memory and computational demands of large…

Global: Single-Agent RL Model Shows Promise for Regional Traffic Signal CoordinationResearchers have introduced a single-agent reinforcement learning (RL) framework designed to manage traffic signals across extensive urban regions, aiming to…