Simplifying the latest STEM breakthroughs for all.
Highlights: Introduces Focus, a Streaming Concentration Architecture for Vision-Language Models (VLMs). Delivers 2.4x faster inference and 3.3x energy reduction compared to existing accelerators. Utilizes a hierarchical compression strategy across semantic,…
Highlights: Researchers develop a multimodal machine learning tool called the Early Warning Index (EWI) to predict patient deterioration. The system integrates structured and unstructured hospital data to forecast ICU admissions,…
Highlights: New framework for dynamic, space-based laser constellations developed. Introduces the Reconfigurable Laser-to-Debris Engagement Scheduling Problem (R-L2D-ESP). Uses optimized orbital reconfiguration to enhance debris removal capacity. Receding horizon technique increases…
Highlights: New Python library ‘gridfm-datakit-v1’ enables scalable and realistic generation of Power Flow (PF) and Optimal Power Flow (OPF) datasets. Developed by a global team of 14 researchers including Alban…
Highlights: New configurable γ photon spectrometer developed for high-precision radioguided tumor resection. CMOS-based integrated circuit measuring 9.9 mm² enables sub-keV energy resolution of single γ photons. Innovative low-power energy spectrometry…
Highlights: Introduces TimeLens, a systematic rethinking of video temporal grounding (VTG) using multimodal large language models (MLLMs). Launches TimeLens-Bench, a high-quality benchmark featuring re-annotated datasets for fairer evaluation. Presents TimeLens-100K,…
Highlights: Introduces a new framework called Collaborative Causal Sensemaking (CCS) for human‑AI decision support. Developed by researchers Raunak Jain and Mudita Khurana. Aims to close the ‘complementarity gap’ in human‑AI…
Highlights: New research investigates how reusing crowdsourced knowledge impacts software development. Study uses large-scale data from Stack Overflow and npm repositories. Findings reveal benefits for productivity but highlight challenges in…
Highlights: Achieved an octave-spanning soliton microcomb with over 50% pump-to-comb conversion efficiency. Overcame the long-standing bandwidth-efficiency trade-off in soliton microcombs. Utilized spatial mode coupling within a single microresonator to maintain…
Highlights: Researchers introduce VGG-Flow, a novel method for fine-tuning flow matching generative models. The approach uses value gradient guidance derived from optimal control theory. Demonstrated efficiency and prior preservation on…