AI news & research for practitioners

Today

Compute Core: Local Inference Engines

NVIDIA GeForce RTX 4090 24GB of GDDR6X VRAM

NVIDIA GeForce RTX 4090 24GB of GDDR6X VRAM

Shop on Amazon →
NVIDIA RTX 4080 Super 16GB of VRAM

NVIDIA RTX 4080 Super 16GB of VRAM

Shop on Amazon →
Major news xAI

Elon Musk’s xAI Is Becoming a Cloud Firm

Elon Musk’s xAI is pivoting towards becoming a cloud computing provider, a move highlighted by its recent agreements to supply computing capacity to AI firms Anthropic and Cursor. This shift...

The Information (headlines) partnership 242w

Yesterday 2026-05-06

Notable news xAI

Is xAI a neocloud now?

xAI, the AI venture founded by Elon Musk, is increasingly being viewed as a potential “neocloud” provider, focusing more on constructing data centers than on developing AI models. This shift...

TechCrunch AI other 275w
Major news Snap stocks

Snap’s Ad Revenue Growth Slowed to 3% in First Quarter

Snap Inc. has reported a significant slowdown in its advertising revenue growth, which increased by only 3% in the first quarter, a stark contrast to the 12% overall revenue growth...

The Information (headlines) earnings 223w
Major news NVIDIA stocks

Nvidia Inks Deal To Invest Up To $3.2 Billion in Corning

Nvidia has announced a significant partnership with Corning, committing to invest up to $3.2 billion to establish three new factories in North Carolina and Texas. This collaboration is particularly timely...

The Information (headlines) partnership 219w
Major news DoorDash stocks

DoorDash’s Revenue Growth Lifted By Deliveroo Purchase

DoorDash has announced a 33% increase in revenue, reaching $4 billion, bolstered significantly by its acquisition of Deliveroo last October. While this growth marks a deceleration from the previous quarter,...

The Information (headlines) deal ma 222w
Critical news OpenAI stocks

Recap: The IPO Reckoning

The initial public offering (IPO) landscape is on the brink of a significant transformation, with SpaceX expected to go public as early as June, potentially achieving a valuation exceeding $1...

The Information (headlines) other 266w
Notable research Hugging Face

vLLM V0 to V1: Correctness Before Corrections in RL

Problem This paper addresses the gap in the literature regarding the reliability and correctness of reinforcement learning (RL) models, particularly in the context of large language models (LLMs). The authors...

Hugging Face Blog training methods 469w
Notable news OpenAI

How Elon Musk left OpenAI, according to Greg Brockman

In a revealing account, Greg Brockman, co-founder of OpenAI, sheds light on the tumultuous negotiations that led to Elon Musk’s departure from the organization. This disclosure is particularly significant as...

TechCrunch AI other 250w
Notable research

Syn4D: A Multiview Synthetic 4D Dataset

Problem The paper addresses the significant gap in high-quality datasets for dense 3D reconstruction and tracking of dynamic scenes from monocular video. Existing datasets often lack comprehensive geometric annotations, which...

arXiv cs.CV other 469w arXiv code Zeren Jiang +5
Notable research Hugging Face

Taming Outlier Tokens in Diffusion Transformers

Problem This paper addresses the underexplored issue of outlier tokens in Diffusion Transformers (DiTs) for image generation, particularly in the context of Representation Autoencoder (RAE)-DiT pipelines. Prior research has identified...

arXiv cs.CV efficiency inference 487w arXiv code Xiaoyu Wu +5
Notable research

Implicit Representations of Grammaticality in Language Models

Problem This preprint addresses the gap in understanding whether pretrained language models (LMs) implicitly acquire a distinction between grammaticality and likelihood. While LMs are designed to maximize corpus likelihood, their...

arXiv cs.CL interpretability 456w arXiv code Yingshan Susan Wang +4
Notable research

The First Token Knows: Single-Decode Confidence for Hallucination Detection

Problem This paper addresses the limitations of existing hallucination detection methods in language models, particularly the inefficiencies of self-consistency and semantic self-consistency approaches. Self-consistency relies on generating multiple sampled answers...

arXiv cs.CL evaluation benchmarks 448w arXiv code Mina Gabriel
Notable research

Aes3D: Aesthetic Assessment in 3D Gaussian Splatting

Problem This paper addresses a significant gap in the literature regarding the aesthetic assessment of 3D scenes generated through 3D Gaussian Splatting (3DGS). Existing methods primarily focus on reconstruction fidelity...

arXiv cs.CV evaluation benchmarks 455w arXiv code Chuanzhi Xu +5
Notable research

What Matters in Practical Learned Image Compression

Problem This paper addresses the gap in the literature regarding the development of a perceptual yet practical learned image codec. While traditional codecs have been optimized for compression efficiency, they...

arXiv cs.CV efficiency inference 398w arXiv code Kedar Tatwawadi +5
Notable research

On the Hardness of Junking LLMs

Problem This preprint addresses a significant gap in the understanding of vulnerabilities in large language models (LLMs), specifically focusing on the existence and discoverability of natural backdoors—token sequences that can...

arXiv cs.LG alignment safety 421w arXiv code Marco Rando +1
Notable research

LineRides: Line-Guided Reinforcement Learning for Bicycle Robot Stunts

Problem This paper addresses the challenge of designing effective reward functions for agile robotic maneuvers in reinforcement learning (RL), particularly for novel platforms and extreme stunts where demonstration-based approaches are...

arXiv cs.AI agents robotics 453w arXiv code Seungeun Rho +5
Notable research

Building informative materials datasets beyond targeted objectives

Problem This preprint addresses the gap in materials science data collection methodologies, specifically the tendency of researchers to focus on a limited subset of properties due to specific research interests....

arXiv cs.AI other 466w arXiv code Rafael Espinosa Castañeda +5
Notable research

Proximal Projection for Doubly Sparse Regularized Models

Problem This paper addresses the limitations of existing regularization techniques in high-dimensional regression, particularly when predictors are structured as a Gaussian graphical model. The authors identify a gap in the...

arXiv cs.LG training methods 405w arXiv code Jia Wei He +2
Notable research

Order Matters: Improving Domain Adaptation by Reordering Data

Problem This paper addresses the challenge of domain shift in machine learning, particularly in the context of unsupervised domain adaptation (UDA). The authors highlight that existing methods for minimizing domain...

arXiv cs.LG training methods 406w arXiv code Andrea Napoli +1
news xAI

Creators of Grok, the AI Chatbot - xAI

xAI, the artificial intelligence company founded by Elon Musk, has unveiled Grok, a new AI chatbot designed to compete with established players like ChatGPT and Google Bard. This launch is...

Google News · xAI / Grok other 253w
Major research Hugging Face

The Impossibility Triangle of Long-Context Modeling

Problem This paper addresses a fundamental limitation in long-sequence modeling, specifically the trade-off between efficiency, compactness, and recall capabilities. The authors present a theoretical framework that proves no model can...

arXiv cs.AI theory 439w arXiv code Yan Zhou
Major research

SoK: Robustness in Large Language Models against Jailbreak Attacks

Problem This paper addresses the significant vulnerability of Large Language Models (LLMs) to jailbreak attacks, which exploit adversarial prompts to elicit harmful or unethical outputs. Despite the proliferation of various...

arXiv cs.AI alignment safety 401w arXiv code Feiyue Xu +5
Major research

Misaligned by Reward: Socially Undesirable Preferences in LLMs

Problem This preprint addresses a significant gap in the evaluation of reward models used for aligning large language models (LLMs) with human preferences. Existing benchmarks primarily assess instruction-following capabilities, neglecting...

arXiv cs.CL alignment safety 493w arXiv code Gayane Ghazaryan +1
Major news xAI stocks

Why SpaceX Might Land in Your Mother’s Index Fund

SpaceX’s anticipated IPO next month could significantly alter the landscape of index funds, particularly if proposed changes by S&P Dow Jones Indices are enacted. These changes would potentially allow high-profile...

The Information (headlines) regulation policy 285w
Notable research

Conceptors for Semantic Steering

Problem This paper addresses the limitations of existing activation-based steering methods for large language models (LLMs), which typically reduce concepts to single directional vectors. This reduction neglects the geometric complexity...

arXiv cs.CL theory 451w arXiv code Ilias Triantafyllopoulos +5
Notable research

Why Expert Alignment Is Hard: Evidence from Subjective Evaluation

Problem This preprint addresses the challenge of aligning large language models (LLMs) with expert judgment in subjective evaluation tasks. The authors highlight a gap in understanding how expert evaluations can...

arXiv cs.CL alignment safety 470w arXiv code Tzu-Mi Lin +4
Major news Anthropic stocks

Rising AI Costs Are Becoming a Problem For Even Investors

AI costs are rapidly escalating, impacting even major players like Uber and venture capital firms. As companies increasingly rely on advanced AI models, the financial implications are becoming unsustainable, prompting...

The Information (headlines) other 249w
Major news TSMC stocks

AI boom pushes Samsung to $1T

Samsung has achieved a significant milestone, surpassing a $1 trillion valuation, driven largely by a surge in demand for AI-focused semiconductor chips. This achievement positions Samsung as only the second...

TechCrunch AI other 250w
Notable research

BenCSSmark: Making the Social Sciences Count in LLM Research

Problem This position paper addresses the significant gap in the representation of social science tasks within existing large language model (LLM) benchmarks, which hampers both LLM evaluation and social scientific...

arXiv cs.CL evaluation benchmarks 466w arXiv code Arnault Chatelain +5
Notable research

Anticipating Innovation Using Large Language Models

Problem This preprint addresses the challenge of forecasting innovation, specifically the emergence of new technological combinations, which has significant implications for science and policy. The authors identify a gap in...

arXiv cs.CL foundation models 412w arXiv code Enrico Maria Fenoaltea +4
Notable news Polymarket

Polymarket’s Homecoming Is Shaky and its U.S. CEO Is AWOL

Polymarket, the prediction market that faced a four-year exile from the U.S., is attempting a comeback by acquiring a licensed derivatives and futures exchange. However, the company’s efforts are marred...

The Information (headlines) regulation policy 221w
Notable news DeepSeek

8 China AI winners riding the DeepSeek breakout - NAI500

A new wave of artificial intelligence companies in China is capitalizing on the recent advancements brought by DeepSeek, a powerful AI model that enhances search capabilities across various sectors. This...

Google News · DeepSeek other 241w
Major news Microsoft stocks

Microsoft Earnings, Apple Earnings

Microsoft has introduced a transformative agentic business model that emphasizes user agency and adaptability, while Apple faces challenges due to shortages in memory and chips, despite seeing positive impacts from...

Stratechery (free) earnings 256w
Major news DeepSeek

DeepSeek in talks for $45b funding round - Tech in Asia

DeepSeek, a prominent player in the AI landscape, is reportedly in discussions to secure a staggering $45 billion in its latest funding round. This potential investment underscores the growing confidence...

Google News · DeepSeek funding round 232w
Major news DeepSeek

China’s National AI Fund in Talks to Invest in DeepSeek

China’s National AI Fund is reportedly in discussions to invest in DeepSeek, a prominent AI company backed by the hedge fund High-Flyer. This potential investment could elevate DeepSeek’s valuation to...

The Information (headlines) funding round 242w
Notable news AMD

Peter Sarlin’s QuTwo reaches $380M valuation in angel round

QyTw0, the Finnish AI lab spearheaded by Peter Sarlin, has achieved a valuation of €325 million (around $380 million) following a successful €25 million angel funding round (approximately $29 million)....

TechCrunch AI funding round 250w
Notable research

AI agents may be skilled researchers—but not always honest ones

Problem This preprint addresses the integrity of AI-generated research outputs, specifically focusing on the propensity of AI agents to fabricate data and engage in p-hacking—manipulating statistical analyses to obtain desired...

Science (AI abstracts) agents robotics 430w
Major news AMD stocks

AMD Shares Rise 16% As AI Hardware Growth Accelerates

AMD, a key player in the AI hardware sector, has reported a significant uptick in its revenue growth, projecting an increase to 46% for the current quarter. This marks an...

The Information (headlines) macro ai demand 232w
Notable research

Core of Solar System’s largest moon may still be forming

Problem This paper addresses the gap in understanding the geophysical processes that govern the magnetic fields of celestial bodies, specifically focusing on Ganymede, the largest moon of Jupiter. The authors...

Science (AI abstracts) other 507w
Major news Google

Google in Talks With Blackstone, KKR to Distribute AI Models

Google is currently in negotiations with private equity giants Blackstone and KKR to enable their portfolio companies to leverage Google’s advanced AI models. This move underscores the growing importance of...

The Information (headlines) partnership 229w
Major news Coinbase

Coinbase’s Missed ‘Learning’ on Hiring

Coinbase recently announced a significant workforce reduction, cutting 14% of its employees, a move that CEO Brian Armstrong attributed to the evolving landscape of AI. This decision comes at a...

The Information (headlines) hiring org changes 221w
Notable news OpenAI

Introducing ChatGPT Futures: Class of 2026

OpenAI has unveiled the ChatGPT Futures Class of 2026, a cohort of 26 student innovators harnessing AI to create impactful solutions across various fields. This initiative highlights the growing role...

OpenAI Blog other 260w
Notable news OpenAI

How frontier enterprises are building an AI advantage

OpenAI’s recent B2B Signals research highlights how leading enterprises are leveraging AI to enhance operational efficiency and create sustainable competitive advantages. As businesses increasingly integrate AI technologies, particularly Codex-powered workflows,...

OpenAI Blog other 228w

2026-05-05 2026-05-05

news OpenAI

Brockman Says Elon Musk Does Not Know AI

In a recent courtroom showdown, Greg Brockman, co-founder and President of OpenAI, asserted that Elon Musk lacks a fundamental understanding of artificial intelligence. This statement came during Musk’s ongoing lawsuit...

The Information (headlines) opinion essay 256w
Major news OpenAI

Private Equity’s AI Deals Lighten the Mood at Milken

Private equity firms are making headlines this week with new partnerships involving AI leaders OpenAI and Anthropic, coinciding with the Milken Institute conference. This surge in AI-related deals is particularly...

The Information (headlines) partnership 223w
Major news ServiceNow

ServiceNow Is Putting Up a New Tollgate for AI Agents

ServiceNow has introduced a new charge for customers using AI agents to access data within its applications, joining the ranks of other tech companies like HubSpot and Workday. This development,...

The Information (headlines) regulation policy 260w
news ASML

ASML CEO Christophe Fouquet: No one is coming for us

ASML CEO Christophe Fouquet, who took the helm of the company in 2024, recently expressed confidence in ASML’s market position during an interview ahead of the Milken Institute Global Conference....

TechCrunch AI other 245w
Major news Apple

Apple Looks to Diversify Chip Manufacturing with Intel and Samsung

Apple is actively pursuing partnerships with Intel and Samsung to expand its chip manufacturing capabilities, a strategic shift aimed at reducing reliance on Taiwan Semiconductor Manufacturing Company (TSMC). This development...

The Information (headlines) infrastructure compute 254w
Notable research Meta

Audio-Visual Intelligence in Large Foundation Models

Problem This paper addresses the fragmented state of the literature on Audio-Visual Intelligence (AVI) in the context of large foundation models. Despite significant advancements in unified audio-vision architectures, existing research...

arXiv cs.CV multimodal 441w arXiv code You Qin +5
Major research

UniCorrn: Unified Correspondence Transformer Across 2D and 3D

Problem This paper addresses the lack of a unified model for visual correspondence across multiple modalities: 2D-2D, 2D-3D, and 3D-3D. Current approaches typically employ task-specific architectures, leading to inefficiencies and...

arXiv cs.CV multimodal 379w arXiv code Prajnan Goswami +3
Notable research

Large Language Models are Universal Reasoners for Visual Generation

Problem This paper addresses the “understanding-generation gap” in text-to-image generation systems, particularly those utilizing large language models (LLMs) and diffusion models. Despite advancements in unified architectures that integrate visual understanding...

arXiv cs.CV reasoning 456w arXiv code Sucheng Ren +5
Major research Meta

Redefining AI Red Teaming in the Agentic Era: From Weeks to Hours

Problem This preprint addresses the inefficiencies in current AI red teaming methodologies, which require operators to manually construct workflows for adversarial testing. Existing approaches are labor-intensive, often taking weeks to...

arXiv cs.AI agents robotics 461w arXiv code Raja Sekhar Rao Dheekonda +2
Notable research

Conditional Diffusion Sampling

Problem This paper addresses the challenge of sampling from unnormalized multimodal distributions with limited density evaluations, a significant gap in the literature. The authors note that while Parallel Tempering (PT)...

arXiv cs.LG other 443w arXiv code Francisco M. Castro-Macías +5
Notable research Hugging Face

RD-ViT: Recurrent-Depth Vision Transformer for Semantic Segmentation with Reduced Data Dependence Extending the Recurrent-Depth Transformer Architecture to Dense Prediction

Problem This paper addresses the significant data dependence of Vision Transformers (ViTs) in semantic segmentation tasks, particularly in medical imaging, where labeled datasets are often limited. The authors propose RD-ViT,...

arXiv cs.CV efficiency inference 449w arXiv code Renjie He
Major news xAI

xAI’s Fast, Cheap Data Center Build-Out Has Hidden Costs

SpaceX is highlighting xAI’s rapid and cost-effective data center construction as a key competitive edge ahead of its initial public offering. The company claims that xAI managed to launch its...

The Information (headlines) infrastructure compute 193w
Notable research

Inconsistent Databases and Argumentation Frameworks with Collective Attacks

Problem This paper addresses the gap in understanding the relationship between subset-maximal repairs for inconsistent databases and argumentation frameworks, particularly when integrity constraints (ICs) include denial constraints and local-as-view tuple-generating...

arXiv cs.AI theory 477w arXiv code Yasir Mahmood +3
Notable research

Towards Open World Sound Event Detection

Problem This paper addresses the limitations of conventional Sound Event Detection (SED) systems, which operate under a closed-world assumption. Such systems are inadequate for real-world applications where novel acoustic events...

arXiv cs.AI other 437w arXiv code P. H. Hai +2
Notable research

Magic-Informed Quantum Architecture Search

Problem This paper addresses the gap in quantum circuit design methodologies that effectively leverage nonstabilizerness, or “magic,” as a resource for achieving quantum advantage. The authors propose a novel approach...

arXiv cs.AI other 447w arXiv code Vincenzo Lipardi +3
Major research

PHALAR: Phasors for Learned Musical Audio Representations

Problem This paper addresses the limitations of existing models in stem retrieval, specifically their inability to effectively utilize temporal information. The authors highlight that current approaches often overlook the significance...

arXiv cs.AI evaluation benchmarks 471w arXiv code Davide Marincione +5
Notable research

Exact ReLU realization of tensor-product refinement iterates

Problem This paper addresses a gap in the literature regarding the exact realization of dyadic refinement operators in two dimensions, specifically for continuous piecewise linear functions. Prior work primarily focused...

arXiv cs.LG theory 455w arXiv code Tsogtgerel Gantumur
Notable research

Steer Like the LLM: Activation Steering that Mimics Prompting

Problem This preprint addresses the gap in performance between activation steering methods and prompt-based steering in large language models (LLMs). While both techniques aim to influence model outputs during inference,...

arXiv cs.AI efficiency inference 444w arXiv code Geert Heyman +1
Notable research

Spatiotemporal Convolutions on EEG signal -- A Representation Learning Perspective on Efficient and Explainable EEG Classification with Convolutional Neural Nets

Problem This preprint addresses the limitations of existing EEG classification methodologies that predominantly utilize independent one-dimensional (1D) convolutional layers for spatial and temporal feature extraction. The authors highlight a gap...

arXiv cs.AI efficiency inference 416w arXiv code Laurits Dixen +2
Major news Shopify stocks

Shopify Shares Slide on Predicted Revenue Slowdown

Shopify has announced a projected slowdown in revenue growth for the second quarter, leading to a 9% drop in its share price. The e-commerce platform attributed this downturn to rising...

The Information (headlines) guidance 217w
Notable research

On Adaptivity in Zeroth-Order Optimization

Problem This paper addresses the gap in the effectiveness of adaptive zeroth-order (ZO) optimization methods for fine-tuning large language models (LLMs) under memory constraints. The authors challenge prior assertions that...

arXiv cs.LG training methods 427w arXiv code Hassan Dbouk +3
Notable research

Memory-Efficient Continual Learning with CLIP Models

Problem This paper addresses the challenge of catastrophic forgetting in Contrastive Language-Image Pretraining (CLIP) models when adapting to new tasks in continual learning scenarios. The authors highlight that existing methods...

arXiv cs.LG efficiency inference 441w arXiv code Ryan King +3
Notable research

Reproducing Complex Set-Compositional Information Retrieval

Problem This paper addresses the gap in understanding how current information retrieval (IR) paradigms handle complex set-compositional queries, which involve conjunction, disjunction, and exclusion. The authors highlight that existing retrieval...

arXiv cs.CL evaluation benchmarks 444w arXiv code Vincent Degenhart +4
Notable research

Realizable Bayes-Consistency for General Metric Losses

Problem This paper addresses the gap in understanding strong universal Bayes-consistency in the realizable setting for learning with general metric losses. Previous works primarily focused on binary classification and real-valued...

arXiv cs.LG theory 424w arXiv code Dan Tsir Cohen +2
Major research

TriBench-Ko: Evaluating LLM Risks in Judicial Workflows

Problem This paper addresses a significant gap in the evaluation of large language models (LLMs) within judicial workflows, particularly in the context of Korean legal systems. Existing benchmarks primarily focus...

arXiv cs.CL evaluation benchmarks 499w arXiv code Haesung Lee +5
Major news OpenAI

Greg Brockman’s Rough Day

In a high-stakes courtroom drama, Greg Brockman, co-founder and president of OpenAI, faced intense scrutiny from Elon Musk’s legal team regarding his substantial financial interests in the AI organization. This...

The Information (headlines) regulation policy 259w
Notable research

Segmenting Human-LLM Co-authored Text via Change Point Detection

Problem This preprint addresses the gap in the capability of existing text detectors to segment human-written and LLM-generated text within co-authored documents. Current methodologies typically yield a binary classification for...

arXiv cs.CL other 454w arXiv code Mengchu Li +3
Notable news Amazon

Amazon Could Offer ‘Hybrid Mode’ AI Search on Retail Site

Amazon is exploring the integration of its Rufus AI shopping assistant into its primary search functionality, potentially introducing a “hybrid mode” that combines traditional search results with AI-generated commentary. This...

The Information (headlines) other 241w
Major news Coinbase stocks

Coinbase to Cut 14% Jobs, Citing Market Conditions and AI

Coinbase, the largest cryptocurrency exchange in the United States, has announced plans to lay off 14% of its workforce, equating to approximately 700 employees. This decision comes amid challenging market...

The Information (headlines) hiring org changes 247w
Notable news Amazon

Amazon’s Durability

Amazon has made significant strides in the AI landscape, particularly in the inference era, positioning itself as a formidable player despite earlier setbacks in the training phase. This shift is...

Stratechery (free) other 257w
Major news OpenAI

GPT-5.5 Instant System Card

OpenAI has unveiled the GPT-5.5 Instant System Card, a significant update that enhances the capabilities of its generative AI models. This release is crucial as it aims to improve user...

OpenAI Blog other 232w
Major news OpenAI

GPT-5.5 Instant: smarter, clearer, and more personalized

OpenAI has unveiled GPT-5.5 Instant, an upgraded version of its ChatGPT model, designed to deliver smarter, clearer, and more personalized responses. This release is particularly significant as it addresses previous...

OpenAI Blog model release 219w
Notable news

A blueprint for using AI to strengthen democracy

A new framework has emerged that outlines how artificial intelligence can be harnessed to bolster democratic processes, drawing parallels to historical shifts in information dissemination that have shaped governance. This...

MIT Technology Review opinion essay 248w
Notable news null

Train Your Own LLM from Scratch

A new hands-on workshop invites participants to build their own language model from scratch, utilizing Andrej Karpathy’s nanoGPT framework. This initiative aims to demystify the process of training large language...

Hacker News (AI filtered) other 272w
Major news Palantir stocks

Palantir’s AI Boom; Ryan Cohen’s Chutzpah

Palantir Technologies has reported a remarkable 85% revenue growth in its first-quarter results, raising questions about its dominance in the enterprise software sector. This surge comes at a time when...

The Information (headlines) earnings 234w
Notable news OpenAI

New ways to buy ChatGPT ads

OpenAI has launched a beta self-serve Ads Manager for ChatGPT, introducing cost-per-click (CPC) bidding and improved measurement tools. This move is significant as it allows advertisers to have more control...

OpenAI Blog other 234w
Notable research

Force-free molecular dynamics through autoregressive equivariant networks

Problem This paper addresses the limitations of traditional molecular dynamics (MD) simulations, which rely heavily on force calculations to predict atomic trajectories. These calculations can be computationally expensive and time-consuming,...

Nature Machine Intelligence other 516w
All news → All research →