PEAR: Position-Embedding-Agnostic Attention Re-weighting Enhances Retrieval-Augmented Generation with Zero Inference Overhead

Kavli Affiliate: Feng Wang | First 5 Authors: Tao Tan, Yining Qian, Ang Lv, Hongzhan Lin, Songhao Wu | Summary: Large language models (LLMs) enhanced with retrieval-augmented generation (RAG) have introduced a new paradigm for web search. However, the limited context awareness of LLMs degrades their performance on RAG tasks. Existing methods to enhance context […]


Continue.. PEAR: Position-Embedding-Agnostic Attention Re-weighting Enhances Retrieval-Augmented Generation with Zero Inference Overhead

PEAR: Position-Embedding-Agnostic Attention Re-weighting Enhances Retrieval-Augmented Generation with Zero Inference Overhead

Kavli Affiliate: Feng Wang | First 5 Authors: Tao Tan, Yining Qian, Ang Lv, Hongzhan Lin, Songhao Wu | Summary: Large language models (LLMs) enhanced with retrieval-augmented generation (RAG) have introduced a new paradigm for web search. However, the limited context awareness of LLMs degrades their performance on RAG tasks. Existing methods to enhance context […]


Continue.. PEAR: Position-Embedding-Agnostic Attention Re-weighting Enhances Retrieval-Augmented Generation with Zero Inference Overhead

Direct measurement of terahertz conductivity in a gated monolayer semiconductor

Kavli Affiliate: Feng Wang | First 5 Authors: Su-Di Chen, Qixin Feng, Wenyu Zhao, Ruishi Qi, Zuocheng Zhang | Summary: Two-dimensional semiconductors and their moir’e superlattices have emerged as important platforms for investigating correlated electrons. However, many key properties of these systems, such as the frequency-dependent conductivity, remain experimentally inaccessible because of the mesoscopic sample […]


Continue.. Direct measurement of terahertz conductivity in a gated monolayer semiconductor

The multi-state geometry of shift current and polarization

Kavli Affiliate: Joel E. Moore | First 5 Authors: Alexander Avdoshkin, Johannes Mitscherling, Joel E. Moore, , | Summary: The quantum metric and Berry curvature capture essential properties of non-trivial Bloch states and underpin many fascinating phenomena. However, it becomes increasingly evident that a more comprehensive understanding of quantum state geometry is necessary to explain […]


Continue.. The multi-state geometry of shift current and polarization

Theoretical insights into the role of lattice fluctuations on the excited behavior of lead halide perovskites

Kavli Affiliate: David T. Limmer | First 5 Authors: Yoonjae Park, Rohit Rana, Daniel Chabeda, Eran Rabani, David T. Limmer | Summary: Unravelling the role of charge-lattice interactions on the optoelectronic properties in lead halide perovskites is of great interest due to their unique photophysical properties. While there is broad consensus on the importance of […]


Continue.. Theoretical insights into the role of lattice fluctuations on the excited behavior of lead halide perovskites

jina-embeddings-v3: Multilingual Embeddings With Task LoRA

Kavli Affiliate: Feng Wang | First 5 Authors: Saba Sturua, Isabelle Mohr, Mohammad Kalim Akram, Michael Günther, Bo Wang | Summary: We introduce jina-embeddings-v3, a novel text embedding model with 570 million parameters, achieves state-of-the-art performance on multilingual data and long-context retrieval tasks, supporting context lengths of up to 8192 tokens. The model includes a […]


Continue.. jina-embeddings-v3: Multilingual Embeddings With Task LoRA

Autoregressive + Chain of Thought $simeq$ Recurrent: Recurrence’s Role in Language Models’ Computability and a Revisit of Recurrent Transformer

Kavli Affiliate: Xiang Zhang | First 5 Authors: Xiang Zhang, Muhammad Abdul-Mageed, Laks V. S. Lakshmanan, , | Summary: The Transformer architecture excels in a variety of language modeling tasks, outperforming traditional neural architectures such as RNN and LSTM. This is partially due to its elimination of recurrent connections, which allows for parallel training and […]


Continue.. Autoregressive + Chain of Thought $simeq$ Recurrent: Recurrence’s Role in Language Models’ Computability and a Revisit of Recurrent Transformer

Reversible Electron-Beam Patterning of Colloidal Nanoparticles at Fluid Interfaces

Kavli Affiliate: Naomi S. Ginsberg | First 5 Authors: Jonathan G. Raybin, Ethan J. Dunsworth, Veronica Guo, Naomi S. Ginsberg, | Summary: The directed self-assembly of colloidal nanoparticles (NPs) using external fields guides the formation of sophisticated hierarchical materials but becomes less effective with decreasing particle size. As an alternative, electron-beam-driven assembly offers a potential […]


Continue.. Reversible Electron-Beam Patterning of Colloidal Nanoparticles at Fluid Interfaces

Voltage Mining for (De)lithiation-stabilized Cathodes and a Machine Learning Model for Li-ion Cathode Voltage

Kavli Affiliate: Kristin A. Persson | First 5 Authors: Haoming Howard Li, Qian Chen, Gerbrand Ceder, Kristin A. Persson, | Summary: Advances in lithium-metal anodes have inspired interest in discovery of Li-free cathodes, most of which are natively found in their charged state. This is in contrast to today’s commercial lithium-ion battery cathodes, which are […]


Continue.. Voltage Mining for (De)lithiation-stabilized Cathodes and a Machine Learning Model for Li-ion Cathode Voltage

Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception

Kavli Affiliate: Xiang Zhang | First 5 Authors: Xiang Zhang, Yufei Cui, Chenchen Fu, Weiwei Wu, Zihao Wang | Summary: Real-time object detection is critical for the decision-making process for many real-world applications, such as collision avoidance and path planning in autonomous driving. This work presents an innovative real-time streaming perception method, Transtreaming, which addresses […]


Continue.. Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception