Articles for category: AI Research

March 11, 2025

ikayaniaamirshahzad@gmail.com

Self-Organising Textures

Contents This article is part of the Differentiable Self-organizing Systems Thread, an experimental format collecting invited short articles delving into differentiable self-organizing systems, interspersed with critical commentary from several experts in adjacent fields. Self-classifying MNIST Digits Adversarial Reprogramming of Neural Cellular Automata Neural Cellular Automata (NCA We use NCA to refer to both Neural Cellular Automata and Neural Cellular Automaton.) are capable of learning a diverse set of behaviours: from generating stable, regenerating, static images , to segmenting images , to learning to “self-classify” shapes . The inductive bias imposed by using cellular automata is powerful. A system of individual

March 11, 2025

ikayaniaamirshahzad@gmail.com

Mean Aggregator is More Robust than Robust Aggregators under Label Poisoning Attacks on Distributed Heterogeneous Data

Mean Aggregator is More Robust than Robust Aggregators under Label Poisoning Attacks on Distributed Heterogeneous Data Jie Peng, Weiyu Li, Stefan Vlaski, Qing Ling; 26(27):1−51, 2025. Abstract Robustness to malicious attacks is of paramount importance for distributed learning. Existing works usually consider the classical Byzantine attacks model, which assumes that some workers can send arbitrarily malicious messages to the server and disturb the aggregation steps of the distributed learning process. To defend against such worst-case Byzantine attacks, various robust aggregators have been proposed. They are proven to be effective and much superior to the often-used mean aggregator. In this paper,

March 11, 2025

ikayaniaamirshahzad@gmail.com

SVLearn: a dual-reference machine learning approach enables accurate cross-species genotyping of structural variants

Sequencing data collection Sequencing data used throughout this study were sourced from publicly available databases. A total of 45 individuals were collected, of which 15 human individuals were obtained from the Human Pangenome Reference Consortium35, 15 sheep individuals were derived from a sheep pangenome study22, and 15 cattle individuals were obtained from five different studies (Charolais36, Holstein37, NxB and OxO38, Yunling39, the remaining ten individuals21). Each individual includes PacBio HiFi long reads and 2×150 bp paired-end short reads (Supplementary Table 1). Construction of ground truth sets for SV genotypes Minimap240 (version 2.26) was used to align the PacBio HiFi long reads of

March 11, 2025

ikayaniaamirshahzad@gmail.com

Open Source Automated Interpretability for Sparse Autoencoder Features

Background# Sparse autoencoders recover a diversity of interpretable, monosemantic features, but present an intractable problem of scale to human labelers. We investigate different techniques for generating and scoring arbitrary text explanations of SAE features, and release a open source library to allow people to do research on auto-interpreted features. Key Findings# Open source models generate and evaluate text explanations of SAE features reasonably well, albeit somewhat worse than closed models like Claude 3.5 Sonnet. Explanations found by LLMs are similar to explanations found by humans. Automatically interpreting 1.5M features of GPT-2 with the current pipeline would cost \$1300 in API

March 11, 2025

ikayaniaamirshahzad@gmail.com

Stanford CRFM

The FMTI team conducts a follow-up study that finds developers are more transparent with ample room for improvement. Visit our website for the paper and transparency reports. Foundation models power impactful AI systems today: Google recently announced that all of its products with at least 2 billion users now rely on the Gemini model. The 2024 AI Index shows that developers are investing hundreds of millions of dollars into building their flagship models. As this technology rises in importance, the demand for transparency escalates. Governments recognize this challenge: the US, EU, China, Canada, and G7 all have taken steps to

March 11, 2025

ikayaniaamirshahzad@gmail.com

The Shift from Models to Compound AI Systems – The Berkeley Artificial Intelligence Research Blog

AI caught everyone’s attention in 2023 with Large Language Models (LLMs) that can be instructed to perform general tasks, such as translation or coding, just by prompting. This naturally led to an intense focus on models as the primary ingredient in AI application development, with everyone wondering what capabilities new LLMs will bring. As more developers begin to build using LLMs, however, we believe that this focus is rapidly changing: state-of-the-art AI results are increasingly obtained by compound systems with multiple components, not just monolithic models. For example, Google’s AlphaCode 2 set state-of-the-art results in programming through a carefully engineered

March 11, 2025

ikayaniaamirshahzad@gmail.com

GenCast predicts weather and the risks of extreme conditions with state-of-the-art accuracy

Technologies Published 4 December 2024 Authors Ilan Price and Matthew Willson New AI model advances the prediction of weather uncertainties and risks, delivering faster, more accurate forecasts up to 15 days ahead Weather impacts all of us — shaping our decisions, our safety, and our way of life. As climate change drives more extreme weather events, accurate and trustworthy forecasts are more essential than ever. Yet, weather cannot be predicted perfectly, and forecasts are especially uncertain beyond a few days. Because a perfect weather forecast is not possible, scientists and weather agencies use probabilistic ensemble forecasts, where the model predicts

March 11, 2025

ikayaniaamirshahzad@gmail.com

[2407.13980] Byzantine-tolerant distributed learning of finite mixture models

[Submitted on 19 Jul 2024 (v1), last revised 10 Mar 2025 (this version, v2)] View a PDF of the paper titled Byzantine-tolerant distributed learning of finite mixture models, by Qiong Zhang and 2 other authors View PDF HTML (experimental) Abstract:Traditional statistical methods need to be updated to work with modern distributed data storage paradigms. A common approach is the split-and-conquer framework, which involves learning models on local machines and averaging their parameter estimates. However, this does not work for the important problem of learning finite mixture models, because subpopulation indices on each local machine may be arbitrarily permuted (the “label

March 11, 2025

ikayaniaamirshahzad@gmail.com

Fully Automatically Evaluating LMMs from the Text-to-Image Generation Perspective

[Submitted on 21 Nov 2024 (v1), last revised 8 Mar 2025 (this version, v2)] View a PDF of the paper titled MMGenBench: Fully Automatically Evaluating LMMs from the Text-to-Image Generation Perspective, by Hailang Huang and 6 other authors View PDF HTML (experimental) Abstract:Large Multimodal Models (LMMs) demonstrate impressive capabilities. However, current benchmarks predominantly focus on image comprehension in specific domains, and these benchmarks are labor-intensive to construct. Moreover, their answers tend to be brief, making it difficult to assess the ability of LMMs to generate detailed descriptions of images. To address these limitations, we propose the MMGenBench-Pipeline, a straightforward and

March 11, 2025

ikayaniaamirshahzad@gmail.com

Data-Efficient Video-LLM with Text-to-Image Augmentation

[Submitted on 29 Nov 2024 (v1), last revised 10 Mar 2025 (this version, v3)] Authors:Shukang Yin, Chaoyou Fu, Sirui Zhao, Yunhang Shen, Chunjiang Ge, Yan Yang, Zuwei Long, Yuhan Dai, Yongdong Luo, Haoyu Cao, Tong Xu, Xing Sun, Caifeng Shan, Ran He, Enhong Chen View a PDF of the paper titled Sparrow: Data-Efficient Video-LLM with Text-to-Image Augmentation, by Shukang Yin and 14 other authors View PDF HTML (experimental) Abstract:Recent years have witnessed the success of Multimodal Large Language Models (MLLMs) in the vision understanding domain. The success of these models can largely be attributed to the dominant scaling law, which