Articles for category: AI Research

March 15, 2025

ikayaniaamirshahzad@gmail.com

Google at APS 2024

New circuits and an open source decoder for the color codePresenter: Craig GidneyAuthors: Craig Gidney, Cody JonesSession S51: Quantum Error Correction Code Performance and Implementation IILink to Paper Performing Hartree-Fock many-body physics calculations with large language modelsPresenter: Eun-Ah KimAuthors: Eun-Ah Kim, Haining Pan, Nayantara Mudur, William Taranto, Subhashini Venugopalan, Yasaman Bahri, Michael P BrennerSession S18: Data Science, AI and Machine Learning in Physics I New methods for reducing resource overhead in the surface codePresenter: Michael NewmanAuthors: Craig M Gidney, Michael Newman, Peter Brooks, Cody JonesSession S51: Quantum Error Correction Code Performance and Implementation IILink to Paper Challenges and opportunities for

March 15, 2025

ikayaniaamirshahzad@gmail.com

Communicating with Interactive Articles

Computing has changed how people communicate. The transmission of news, messages, and ideas is instant. Anyone’s voice can be heard. In fact, access to digital communication technologies such as the Internet is so fundamental to daily life that their disruption by government is condemned by the United Nations Human Rights Council . But while the technology to distribute our ideas has grown in leaps and bounds, the interfaces have remained largely the same. Parallel to the development of the internet, researchers like Alan Kay and Douglas Engelbart worked to build technology that would empower individuals and enhance cognition. Kay imagined

March 15, 2025

ikayaniaamirshahzad@gmail.com

Sliced-Wasserstein Distances and Flows on Cartan-Hadamard Manifolds

Sliced-Wasserstein Distances and Flows on Cartan-Hadamard Manifolds Clément Bonet, Lucas Drumetz, Nicolas Courty; 26(32):1−76, 2025. Abstract While many Machine Learning methods have been developed or transposed on Riemannian manifolds to tackle data with known non-Euclidean geometry, Optimal Transport (OT) methods on such spaces have not received much attention. The main OT tool on these spaces is the Wasserstein distance, which suffers from a heavy computational burden. On Euclidean spaces, a popular alternative is the Sliced-Wasserstein distance, which leverages a closed-form solution of the Wasserstein distance in one dimension, but which is not readily available on manifolds. In this work, we

March 15, 2025

ikayaniaamirshahzad@gmail.com

Feature selection methods affect the performance of scRNA-seq data integration and querying

Our study follows a standard benchmark design, consisting of test datasets, feature selection methods to be evaluated and metrics for measuring performance (Extended Data Fig. 1). The complete benchmarking pipeline is implemented as a Nextflow50 workflow (Extended Data Fig. 2) available from GitHub51 and archived on Zenodo52. Summaries of the specific methods, metrics, datasets and processing steps are provided in the following sections. Please refer to the supplementary methods, pipeline code, original publications and package documentation for further information. Evaluated methods We selected a range of feature selection methods covering approaches from standard analysis workflows and alternative methods proposed for

March 15, 2025

ikayaniaamirshahzad@gmail.com

The Foundation Model Development Cheatsheet

The pace of foundation model releases and progress has continued to grow rapidly over the past few years, with many new models released from organizations of all kinds worldwide. In addition to releasing models themselves, it’s also important to make the tools to create these models – large-scale training libraries, data processing and creation tooling, and more – widely available. In April 2023 we released the Pythia model suite, the first LLMs with a fully released and reproducible technical pipeline from start to finish. We are excited to see other organizations following suit, with the LLM360 project releasing Amber later

March 15, 2025

ikayaniaamirshahzad@gmail.com

Stanford CRFM

Introducing HELM Capabilities, a benchmark that evaluates language models across a curated set of key capabilities, providing a comparison of their strengths and weaknesses Evaluating language models is a dynamic and critical process as models continue to improve rapidly. Understanding their strengths and weaknesses is essential for external users to determine which models suits their needs. Two years ago, we introduced the Holistic Evaluation of Language Models (HELM) as a framework to assess language models. Using the HELM framework, we have published a series of transparent and reproducible leaderboards, including general-purpose leaderboards for evaluating core capabilities (HELM Classic, HELM Lite),

March 15, 2025

ikayaniaamirshahzad@gmail.com

A new era of discovery

AI is revolutionizing the landscape of scientific research, enabling advancements at a pace that was once unimaginable — from accelerating drug discovery to designing new materials for clean energy technologies. The AI for Science Forum — co-hosted by Google DeepMind and the Royal Society — brought together the scientific community, policymakers, and industry leaders to explore the transformative potential of AI to drive scientific breakthroughs, address the world’s most pressing challenges, and lead to a new era of discovery. Source link

March 15, 2025

ikayaniaamirshahzad@gmail.com

[2411.00113] A Geometric Framework for Understanding Memorization in Generative Models

[Submitted on 31 Oct 2024 (v1), last revised 12 Mar 2025 (this version, v2)] View a PDF of the paper titled A Geometric Framework for Understanding Memorization in Generative Models, by Brendan Leigh Ross and 7 other authors View PDF HTML (experimental) Abstract:As deep generative models have progressed, recent work has shown them to be capable of memorizing and reproducing training datapoints when deployed. These findings call into question the usability of generative models, especially in light of the legal and privacy risks brought about by memorization. To better understand this phenomenon, we propose the manifold memorization hypothesis (MMH), a

March 15, 2025

ikayaniaamirshahzad@gmail.com

Adaptive Multi-Modal Multi-View Fusion for 3D Human Body Reconstruction

[Submitted on 7 Sep 2024 (v1), last revised 13 Mar 2025 (this version, v3)] View a PDF of the paper titled AdaptiveFusion: Adaptive Multi-Modal Multi-View Fusion for 3D Human Body Reconstruction, by Anjun Chen and 7 other authors View PDF HTML (experimental) Abstract:Recent advancements in sensor technology and deep learning have led to significant progress in 3D human body reconstruction. However, most existing approaches rely on data from a specific sensor, which can be unreliable due to the inherent limitations of individual sensing modalities. Additionally, existing multi-modal fusion methods generally require customized designs based on the specific sensor combinations or

March 15, 2025

ikayaniaamirshahzad@gmail.com

A New Benchmark for Assessing Hallucination in Medical Large Language Models

[Submitted on 25 Dec 2024 (v1), last revised 13 Mar 2025 (this version, v3)] View a PDF of the paper titled MedHallBench: A New Benchmark for Assessing Hallucination in Medical Large Language Models, by Kaiwen Zuo and 1 other authors View PDF HTML (experimental) Abstract:Medical Large Language Models (MLLMs) have demonstrated potential in healthcare applications, yet their propensity for hallucinations — generating medically implausible or inaccurate information — presents substantial risks to patient care. This paper introduces MedHallBench, a comprehensive benchmark framework for evaluating and mitigating hallucinations in MLLMs. Our methodology integrates expert-validated medical case scenarios with established medical databases