Rethinking Key-Value Cache Compression Techniques for Large Language Model Serving

Kavli Affiliate: Wei Gao | First 5 Authors: Wei Gao, Xinyu Zhou, Peng Sun, Tianwei Zhang, Yonggang Wen | Summary: Key-Value cache (texttt{KV} texttt{cache}) compression has emerged as a promising technique to optimize Large Language Model (LLM) serving. It primarily decreases the memory consumption of texttt{KV} texttt{cache} to reduce the computation cost. Despite the development […]


Continue.. Rethinking Key-Value Cache Compression Techniques for Large Language Model Serving

The Potential of the SPHEREx Mission for Characterizing PAH 3.3 μm Emission in Nearby Galaxies

Kavli Affiliate: James J. Bock | First 5 Authors: Edward Zhang, Andreas L. Faisst, Brendan Crill, Hanae Inami, Thomas Lai | Summary: Dust is a key galaxy component together with gas, stars, and central supermassive black holes, playing a crucial role in stellar and galaxy evolution. Hence, it is critical to understand galaxies’ dust content […]


Continue.. The Potential of the SPHEREx Mission for Characterizing PAH 3.3 μm Emission in Nearby Galaxies

Structure and Melting of Fe, MgO, SiO2, and MgSiO3 in Planets: Database, Inversion, and Phase Diagram

Kavli Affiliate: Paul D. Asimow | First 5 Authors: Junjie Dong, Gabriel-Darius Mardaru, Paul D. Asimow, Lars P. Stixrude, Rebecca A. Fischer | Summary: We present globally inverted pressure-temperature (P-T) phase diagrams up to 5,000 GPa for four fundamental planetary materials, Fe, MgO, SiO2, and MgSiO3, derived from logistic regression and supervised learning, together with […]


Continue.. Structure and Melting of Fe, MgO, SiO2, and MgSiO3 in Planets: Database, Inversion, and Phase Diagram

UniPCGC: Towards Practical Point Cloud Geometry Compression via an Efficient Unified Approach

Kavli Affiliate: Wei Gao | First 5 Authors: Kangli Wang, Wei Gao, , , | Summary: Learning-based point cloud compression methods have made significant progress in terms of performance. However, these methods still encounter challenges including high complexity, limited compression modes, and a lack of support for variable rate, which restrict the practical application of […]


Continue.. UniPCGC: Towards Practical Point Cloud Geometry Compression via an Efficient Unified Approach

NMPC-based Unified Posture Manipulation and Thrust Vectoring for Fault Recovery

Kavli Affiliate: Morteza Gharib | First 5 Authors: Adarsh Salagame, Shashwat Pandya, Ioannis Mandralis, Eric Sihite, Alireza Ramezani | Summary: Multi-rotors face significant risks, as actuator failures at high altitudes can easily result in a crash and the robot’s destruction. Therefore, rapid fault recovery in the event of an actuator failure is necessary for the […]


Continue.. NMPC-based Unified Posture Manipulation and Thrust Vectoring for Fault Recovery

Learning Orientation Field for OSM-Guided Autonomous Navigation

Kavli Affiliate: Wei Gao | First 5 Authors: Yuming Huang, Wei Gao, Zhiyuan Zhang, Maani Ghaffari, Dezhen Song | Summary: OpenStreetMap (OSM) has gained popularity recently in autonomous navigation due to its public accessibility, lower maintenance costs, and broader geographical coverage. However, existing methods often struggle with noisy OSM data and incomplete sensor observations, leading […]


Continue.. Learning Orientation Field for OSM-Guided Autonomous Navigation

SPPO:Efficient Long-sequence LLM Training via Adaptive Sequence Pipeline Parallel Offloading

Kavli Affiliate: Wei Gao | First 5 Authors: Qiaoling Chen, Shenggui Li, Wei Gao, Peng Sun, Yonggang Wen | Summary: In recent years, Large Language Models (LLMs) have exhibited remarkable capabilities, driving advancements in real-world applications. However, training LLMs on increasingly long input sequences imposes significant challenges due to high GPU memory and computational demands. […]


Continue.. SPPO:Efficient Long-sequence LLM Training via Adaptive Sequence Pipeline Parallel Offloading

The role of effective mass and long-range interactions in the band-gap renormalization of photo-excited semiconductors

Kavli Affiliate: Scott K. Cushing | First 5 Authors: Cian C. Reeves, Scott K. Cushing, Vojtech Vlcek, , | Summary: Understanding how to control changes in electronic structure and related dynamical renormalizations by external driving fields is the key for understanding ultrafast spectroscopy and applications in electronics. Here we focus on the band-gap’s modulation by […]


Continue.. The role of effective mass and long-range interactions in the band-gap renormalization of photo-excited semiconductors

High-Precision Transformer-Based Visual Servoing for Humanoid Robots in Aligning Tiny Objects

Kavli Affiliate: Wei Gao | First 5 Authors: Jialong Xue, Wei Gao, Yu Wang, Chao Ji, Dongdong Zhao | Summary: High-precision tiny object alignment remains a common and critical challenge for humanoid robots in real-world. To address this problem, this paper proposes a vision-based framework for precisely estimating and controlling the relative position between a […]


Continue.. High-Precision Transformer-Based Visual Servoing for Humanoid Robots in Aligning Tiny Objects

ATMO: An Aerially Transforming Morphobot for Dynamic Ground-Aerial Transition

Kavli Affiliate: Morteza Gharib | First 5 Authors: Ioannis Mandralis, Reza Nemovi, Alireza Ramezani, Richard M. Murray, Morteza Gharib | Summary: Designing ground-aerial robots is challenging due to the increased actuation requirements which can lead to added weight and reduced locomotion efficiency. Morphobots mitigate this by combining actuators into multi-functional groups and leveraging ground transformation […]


Continue.. ATMO: An Aerially Transforming Morphobot for Dynamic Ground-Aerial Transition