🗄️ News Archive

Gigabyte's Aorus RTX 5090: extreme cooling to surpass Asus' ROG Matrix

Gigabyte aims for the top with its new Aorus RTX 5090, directly challenging the Asus ROG Matrix. The card promises superior performance thanks to a cutting-edge cooling system, combining superconducting heatpipes and a double flow-through design to maximize heat dissipation.

Sega co-founder David Rosen passed away on Christmas Day at 95

David Rosen, one of the co-founders of Sega, has passed away. A former U.S. airman, Rosen helped establish Sega after his service ended in Japan. He was 95 years old. Sega played a crucial role in the evolution of the video game industry, competing with Nintendo in the 1990s.

Asus launches new ROG Zephyrus laptops with RTX 5090 GPU

Asus has unveiled two new ROG Zephyrus laptops at CES, available with 14-inch and 16-inch displays. They feature Intel Core Ultra 300 or AMD Ryzen AI CPUs, and up to an RTX 5090 GPU. The new models promise high performance for gaming and content creation, combining power and portability in a sleek and compact design.

3 best secure container images for modern applications

The security of container images is crucial for modern applications. Echo, Google Distroless, and Ubuntu Containers offer different approaches to reduce vulnerabilities and improve reliability. The choice depends on the specific needs of the organization, considering factors such as vulnerability management, completeness, and ecosystem compatibility.

Nvidia CEO confirms Vera Rubin NVL72 is now in production

Nvidia CEO Jensen Huang announced during his keynote speech at CES that Vera Rubin NVL72 is now officially in production. The new datacenter AI platform offers up to 5x the performance of Blackwell counterparts.

Why AI predictions are so hard

Predictions about artificial intelligence (AI) have become more complex due to key uncertainties. The future of large language models (LLMs) is undefined, public opinion is predominantly negative towards AI, and lawmakers' responses are mixed. Despite AI's progress in science, doubts remain about its effectiveness in other sectors, making it difficult to predict its future impact.

Multi-Dimensional Prompt Chaining Boosts Open-Domain Dialogue in SLMs

A new multi-dimensional prompt-chaining framework aims to enhance the dialogue quality of small language models (SLMs) in open-domain settings. By integrating Naturalness, Coherence, and Engagingness dimensions, the system allows TinyLlama and Llama-2-7B to rival much larger models like Llama-2-70B and GPT-3.5 Turbo.

HyperJoin: AI and Hypergraphs for Joinable Table Discovery in Data Lakes

A new framework, HyperJoin, leverages large language models (LLMs) and hypergraphs to improve the discovery of joinable tables in data lakes. The system models tables as hypergraphs, formulates discovery as link prediction, and uses a hierarchical interaction network for more expressive representations, increasing precision and recall compared to existing solutions.

Intention Collapse: Measuring Intentions in Language Models

A new study introduces metrics to analyze how language models compress intentions into token sequences. Researchers defined three model-agnostic metrics – intention entropy, effective dimensionality, and latent knowledge recoverability – and conducted experiments on a 4-bit Mistral 7B model to evaluate the effectiveness of "chain of thought" in reducing entropy and improving accuracy.

New Approach for Efficient Inference with Memory-Constrained AI Agents

A new study introduces "compressed query delegation" (CQD) to enhance the reasoning abilities of memory-constrained AI agents. The method compresses latent reasoning states, delegates queries to external oracles, and updates states via Riemannian optimization. Results show improvements over traditional methods in complex tasks.

LLMs for Sociological Simulations: A Qualitative Laboratory

A new study explores the use of Large Language Models (LLMs) to simulate personas and generate qualitative hypotheses in the sociological field. The method offers advantages over traditional surveys and rule-based models, opening new avenues for social research and understanding reactions to specific stimuli.

JEPA: Value-guided action planning with world models

A new study explores how to improve action planning in Joint-Embedded Predictive Architectures (JEPA) models, by modeling environmental dynamics through representations and self-supervised prediction objectives. The proposed method shapes the representation space, approximating the goal-conditioned value function with a distance between states, significantly improving planning performance in control tasks.

SLO-Conditioned Action Routing for Retrieval-Augmented Generation

A new study explores per-query control in Retrieval-Augmented Generation (RAG) systems, modeling the choice between different retrieval depths, generation modes, and query refusal. The goal is to satisfy service-level objectives (SLOs) such as cost, refusal rate, and hallucination risk. The results highlight the importance of careful evaluation of learned policies and potential failure modes.

IM-PINN: Geometric Simulations with Neural Networks for Differential Equations

A novel geometric deep learning framework, called IM-PINN, promises to solve partial differential equations on complex Riemannian manifolds without the use of meshes. The system is based on neural networks and aims to overcome the limitations of traditional methods, offering greater accuracy and efficiency in computation.

Artificial intelligence for shrimp disease diagnosis

A new study explores the use of deep learning to automatically classify shrimp diseases, crucial for sustainable production. Using a dataset of 1,149 images and several pre-trained models, researchers achieved 96.88% accuracy with ConvNeXt-Tiny, opening new perspectives for monitoring and managing diseases in the aquaculture sector.

Horizon Reduction as Information Loss in Offline Reinforcement Learning

A new study analyzes Horizon Reduction (HR) in offline Reinforcement Learning (RL), a technique used to improve stability and scalability. The research demonstrates that HR can cause a fundamental and irrecoverable loss of information, making optimal policies indistinguishable from suboptimal ones, even with infinite data. Three structural failure modes are identified, highlighting the intrinsic limitations of HR.

Energy-Aware Routing for Large Reasoning Models

A new study explores how to reduce the energy consumption of large reasoning models (LRMs). The key is to balance the mean energy provisioning and stochastic fluctuations, avoiding waste. Variance-aware routing and dispatch policies based on training-compute and inference-compute scaling laws are crucial for energy efficiency.

CogCanvas: Enhanced Memory for Long LLM Conversations

CogCanvas is a new framework that enhances memory management in large language models (LLMs) during extended conversations. Unlike traditional methods that truncate or summarize information, CogCanvas extracts key elements such as decisions and facts, organizing them into a temporal graph. Tests demonstrate a significant improvement in accuracy, especially in temporal and causal reasoning, compared to other techniques like RAG and GraphRAG.

Agentic AI for Autonomous, Explainable, and Real-Time Credit Risk Decision-Making

A new study explores the use of Agentic AI systems to automate and make credit risk decisions more transparent. The proposed system aims to overcome the limitations of traditional machine learning models, offering greater adaptability and situational awareness, while addressing challenges such as model drift and regulatory uncertainties.

MathLedger: A Verifiable Learning Substrate with Ledger-Attested Feedback

MathLedger, a system integrating formal verification, cryptographic attestation, and learning dynamics for more transparent and reliable AI systems, has been introduced. The prototype implements Reflexive Formal Learning (RFL), a symbolic approach to learning based on verifier outcomes rather than statistical loss. Initial tests validate its measurement and governance infrastructure, paving the way for verifiable learning systems at scale.

Semantic Alignment of Multilingual Knowledge Graphs via Contextualized Vector Projections

A new system for cross-lingual ontology alignment leverages embedding-based cosine similarity matching. The system enriches ontology entities with contextual descriptions and uses a fine-tuned transformer-based multilingual model to generate better embeddings. Evaluated on the OAEI-2022 multifarm track, the system achieved an F1 score of 71%, a 16% increase from the best baseline score.

AMD unveils Instinct MI400X AI accelerators

At CES, AMD unveiled the full MI400X series of AI accelerators, including the Instinct MI430X, MI440X, and MI455X models. The new Helios architecture aims to meet the diverse infrastructure and customer requirements in the field of artificial intelligence and high-performance computing (HPC).

AMD's Ryzen AI 400 series: first Copilot+ desktop CPU from Team Red

AMD refreshes Zen 5 APUs and Strix Halo with the Ryzen AI 400 series, featuring the first Copilot+ desktop CPU. The new Gorgon Point series is primarily driven by a clock speed bump, featuring similar silicio as the previous generation.

AMD's CES 2026 Keynote: Gorgon Point, X3D updates, and AI

AMD CEO Lisa Su will take to the stage at CES 2026 to outline the company's latest advances. The keynote will focus on Gorgon Point, X3D technology updates, and artificial intelligence applications. A deep dive into AMD's future strategies in these key areas is expected.

Hilton Hotel That Refused DHS Reservations Backpedals

A Hilton branded hotel originally declined to host guests because they were with the Department of Homeland Security (DHS). The hotel chain later apologized, stating that it does not tolerate discrimination. The hotel is independently operated, and according to Hilton, the actions taken do not reflect the group's values. The episode has raised a debate about the obligation to provide services to everyone.

Jensen Huang Says Nvidia’s New Vera Rubin Chips Are in ‘Full Production’

Jensen Huang announced that Nvidia's new Vera Rubin chips are in full production. These chips promise to sharply cut the cost of training and running AI models, strengthening the appeal of Nvidia's integrated computing platform. A crucial step forward in democratizing access to AI and accelerating innovation in the sector.

Microsoft’s Nadella: Stop Thinking of AI as a Problem

Microsoft CEO Satya Nadella urges a shift in perspective, viewing AI not as a job killer but as a helpful assistant. New data for 2026 suggests this vision may be accurate, pointing towards a future of human-AI collaboration.

Acer refreshes Swift laptops with Panther Lake at CES 2026

Acer announced at CES 2026 an update to its Swift, Swift Edge, and Swift Go laptop lineup. The new models will integrate the latest Intel processors, offering superior performance. Pricing has not yet been announced.

Nvidia wants to be the Android of generalist robotics

At CES 2026, Nvidia unveiled a full-stack robotics ecosystem, including foundation models, simulation tools, and hardware. The company aims to become the default platform for generalist robotics, similar to Android in the mobile world.

Nvidia Vera Rubin NVL72: AI supercomputer promises 5x performance boost

At CES 2026, Nvidia unveiled Vera Rubin NVL72, a next-generation rack-scale AI accelerator. It promises 5x greater inference performance and 10x lower cost per token than the last-gen Grace Blackwell NVL72. Shipping is expected in the second half of 2026.

Nvidia launches powerful new Rubin chip architecture

Nvidia CEO Jensen Huang officially launched the company’s new Rubin computing architecture, which he described as the state of the art in AI computing. Nvidia continues to heavily invest in research and development to maintain its leadership in the GPU and AI solutions market.

Grok's AI Creates Sexualized Images from Uploaded Photos

The integration of Grok AI into X has led to the creation of non-consensual sexualized images, often from photos of women, celebrities, and even minors. The lack of content moderation on the platform exacerbates the problem, raising ethical concerns and the spread of disinformation.

Nvidia CES 2026 keynote: Jensen Huang unveils AI roadmap

Nvidia CEO Jensen Huang will deliver a keynote at CES 2026 in Las Vegas. Following a record year for the company, major announcements on the future of artificial intelligence are expected. No new consumer GPUs are anticipated.

Nvidia launches Alpamayo, open AI models that allow autonomous vehicles to ‘think like a human’

Nvidia unveiled Alpamayo at CES 2026, which includes a reasoning vision language action model that allows an autonomous vehicle to think more like a human and provide chain-of-thought reasoning.

Nvidia: No New GPUs at CES for the First Time in 5 Years

Nvidia has announced that it will not unveil any new GPUs at CES 2026, breaking a five-year streak. The company will instead focus on AI developments. This decision marks a shift in strategy for Nvidia, which has consistently used CES to unveil its latest innovations in graphics cards.

X blames users for Grok-generated CSAM; no fixes announced

X is blaming users for generating child sexual abuse material (CSAM) with Grok. The company has not announced any fixes to the system, but threatens suspensions and legal action for those who abuse the tool.

Qualcomm expands Snapdragon on Windows with X2 Plus – 10-core ARM CPU

Qualcomm is expanding its second generation of Windows chips with the X2 Plus lineup, which claims a 35% jump in single-core and a 17% increase in multi-threaded performance. This represents a significant evolution for Windows devices based on ARM architecture.

Researchers Are Hunting America for Hidden Datacenters

A research team at Epoch AI is mapping the growth of America's datacenters using open-source intelligence. They analyze satellite imagery and local legal documents to create an interactive map detailing costs, power output, and owners, providing crucial information about these expanding infrastructures.

DoorDash bans driver for allegedly faking delivery with AI

DoorDash has confirmed that it banned a driver who allegedly used an AI-generated photo to fake a delivery. The incident went viral on social media, raising questions about the misuse of AI and the controls of delivery platforms.

CES 2026: Industry Focuses on Artificial Intelligence

La conferenza CES 2026 ha visto l'annuncio di nuovi sistemi di gestione dei dati, GPU e CPU che utilizzano l'intelligenza artificiale per migliorare la sicurezza e le prestazioni.

Chip scarcity assaults auto industry amid the worsening Nexperia and DRAM crisis — fragile sector rocked by undersupply, which may end worse than the 2021 shortage

The automotive industry faces a fresh chip crisis due to DRAM demand and the rapid scale-up of AI infrastructure.

CES award out MSI's monstrous 1600W RTX 5090 Lightning GPU

MSI has confirmed its new flagship GPU, the RTX 5090 Lightning, at CES with a series of innovations that make it an extremely interesting option for extreme overclocking enthusiasts.

The Data Center Boom Is Concentrated in the U.S.

Gli Stati Uniti hanno più della metà delle nuove strutture previste per i data center, ma la capacità del sistema elettrico è una preoccupazione importante.

AI Deepfakes Are Impersonating Pastors to Try to Scam Their Congregations

Le comunità religiose negli Stati Uniti sono state colpite da dipicature AI di i loro leader che condividono discorsi incendiari e chiedono donazioni.

Everything Is AI at CES 2026: What Matters Is How You Use It

Il CES 2026 è stato un evento che ha segnato l'inizio di una nuova era per la tecnicia: l'era dell'intelligenza artificiale. La chatbot integrata e l'intelligenza del machine sono ormai diventate standard in consumer tech, ma se le aziende vogliono vincere in questo contesto, devono concentrarsi sulla qualità dell'esperienza utente.

A Physical Theory of Intelligence

Recent scientific research has led to a new theory of intelligence based on the understanding of information physics. The author presents a framework called Conservation-Congruent Encoding (CCE) that links intelligence to physical laws.

MetaJuLS: a novel approach to pushing structured inference in large language models

MetaJuLS is a meta-learning solution that introduces an universal approach to constraint propagation for structured inference in large language models. This approach achieves 1.5-2.0x speedups over GPU-optimized baselines while maintaining within 0.2% accuracy of state-of-the-art parsers.

Evaluating Anomaly Detectors for Simulated Highly Imbalanced Industrial Classification Problems

A comprehensive evaluation of anomaly detection algorithms using a problem-agnostic simulated dataset that reflects real-world engineering constraints.

Introduction to Reasoning in Action for Large Language Models

Un nuovo approccio per integrare la ricerca e la ragione negli LLMs. Il metodo introduce una strategia di recupero del sapere che si concentra sulla struttura logica delle conversazioni, migliorando così il rendimento dei modelli.

RIMRULE: A New Approach for Adapting Language Models to Specific Tools

Researchers have proposed a new method for adapting language models to specific tools, called RIMRULE. This approach uses dynamic rule injection and an MDL objective to favor generality and conciseness of rules.

Knowledge Distillation for Temporal Knowledge Graph Reasoning with Large Language Models

The decline of AI system performance is a fundamental problem to improve efficiency and reliability. Temporal knowledge graph reasoning is crucial for future AI applications.

Chain-of-Legal-Thought Evaluation for Patent Description

The Pat-DEVAL team has presented a new evaluation framework for patent descriptions, called Chain-of-Legal-Thought Evaluation. This approach uses large language models to evaluate the structural coherence and statutory compliance of patent descriptions.

Exploration in the Limit

A research group has introduced a new form of optimal identification with limited error control, overcoming the limitations of existing methodologies.

The Trojan in the Vocabulary: Stealthy Sabotage of LLM Composition

A new attack discovered on the LLM composition system may compromise model security.

A Multi-Algorithm Approach for Operational Human Resources Workload Balancing in a Last Mile Urban Delivery System

La gestione del carico di lavoro operativo delle risorse umane è critica nel sistema di consegna last mile. L'approccio multi-algoritmo proposto utilizza una combinazione di considerazioni di distanza e carico di lavoro per ottimizzare l'allocazione delle consegne ai lavoratori.

Intel Announces Upcoming Arc B770 GPU, Launch Appears Imminent

Intel has released a new code on GitHub, which suggests the upcoming arrival of the Arc B770 GPU. The card seems to have been tested and supported by an update to add support.

Scientists Discover One of the World’s Oldest Cremations

A team of scientists has discovered an ancient cremation in Africa, dating back over 9,500 years. The event was found at a site in Malawi and represents one of the oldest examples of cremated remains found in a pyre.

Plaud launches new AI pin and desktop meeting notetaker

Plaud has launched a new app that records online meetings and offers a more comprehensive user experience.

Samsung announces new OLED products

Samsung has announced a new display product that promises significant innovations. The company confirmed that the new displays will be showcased at CES 2026 and that the product will feature advanced features such as digital audio reproduction.

Windows 11 at the Bottom of the Classify

A new speed test has put the performance of the latest OS to the test against the five previous versions. The surprise? Windows 8.1, which emerges as a winner in some tests.

Nvidia's RTX 5070: German Retailer Allegedly Rationing Supply

La Germania si trova in una situazione critica nel mercato delle GPU, con un calo significativo della disponibilità dei nuovi modelli. Un supposto fornitore tedesco ha rivelato che le consegne di RTX 5070 saranno limitate a cinque unità per modello, mentre i modelli più potenti sono completamente fuori stock.

US Army tests 20kW vehicle-mounted laser systems for drone defense — updated LOCUST device is now more lethal against drones

La US Army sta testando un nuovo sistema laser per difendere i droni, migliorato rispetto alla versione precedente. Il sistema, sviluppato da AeroVironment Inc., ha una potenza di 20 kW e può essere montato su veicoli.

Disinformation Floods Social Media After Nicolás Maduro's Capture

Disinformation Floods Social Media After Nicolás Maduro's Capture

MSI's RTX 5090 Lightning shatters GPU records before launch — 40-phase VRAM and dual 12V-2x6 connectors turn heads on upcoming overclocking monster

MSI's RTX 5090 Lightning shatters GPU records before launch — 40-phase VRAM and dual 12V-2x6 connectors turn heads on upcoming overclocking monster

La disputa sul Venezuela: ChatGPT in disaccordo con gli Stati Uniti

The New York Times reports that US intelligence agencies invaded Venezuela and captured President Nicolás Maduro. But ChatGPT, the Meta LLM model, disputes this account.

Laser cutting and engraving from a software perspective — master your hardware and optimize your creative output with some software know-how

We take a closer look at the software side of laser cutting and engraving.

OpenAI si prepara a conquistare l'audio con nuovi modelli e hardware

La società di AI OpenAI sta ristrutturando alcuni team per sviluppare prodotti hardware basati su tecnologie audio, con l'obiettivo di migliorare la precisione e la velocità dei modelli. La nuova piattaforma sarà focalizzata sull'audio e si spera che potrà spingere gli utenti a utilizzare più frequentemente l'interfaccia vocale.

Scoperta rivoluzionaria sui resti dell'antenato umano più antico

Scientists Identify Remains of the Earliest Human Ancestor

India orders Musk’s X to fix Grok over “obscene” AI content

India orders Musk’s X to fix Grok over “obscene” AI content

How AI is reshaping work and who gets to do it, according to Mercor’s CEO

Mercor, a three-year-old startup, has become a $10 billion middleman in AI's data gold rush. The company connects AI labs like OpenAI and Anthropic with former employees of Goldman Sachs, McKinsey, and white-shoe law firms, paying them up to $200 an hour to share their industry expertise and train the AI models that could eventually automate their former employers out of business.

Micron si consolida come leader nel mercato di memoria con una sussidio di 318 milioni di dollari

Micron secures $318 million Taiwanese subsidy for HBM R&D as AI memory arms race intensifies — three-year project aims to develop leading-edge, high-performance memory

MSI protegge i cavi della RTX 5090 con nuove power supply

MSI presenta due nuove power supply con 'instant protection' contro la fusione dei cavi RTX 5090.

L'Impero dell'IoT di Nvidia: come stanno andando le sue startup

Nvidia ha investito oltre 100 startup nel campo dell'intelligenza artificiale in due anni. Ecco i suoi maggiori investimenti.

Debian maintainer criticizes email-managed bug tracker as outdated for modern development

A dispute has flared up inside the Debian project after a senior maintainer criticized the distribution’s bug tracking system as outdated and increasingly unworkable for modern software development

In 2026, AI will move from hype to pragmatism

In 2026, here's what you can expect from the AI industry: new architectures, smaller models, world models, reliable agents, physical AI, and products designed for real-world use.

Solana: la piattaforma di blockchain che sta cambiando il gioco per gli AI

La piattaforma Solana sta diventando sempre più popolare tra i progettisti di programmi autonomi (AI) grazie alla sua velocità e stabilità. Tuttavia, questo aumento della domanda sta anche aumentando la minaccia di attacchi maliciosi.

La catena di produzione dei chip si stringe: cosa ci aspetta nel 2026?

L'industria del settore dei chip è ancora sotto l'ombra della crisi globale, e gli analisti prevedono un futuro conservatore per la capacità di produzione.

Reports of the previously-recalled Bambu Lab 3D printer being a fire hazard resurface, new power board component issue identified as the cause — Bambu says the A1's issue has been addressed, affected units replaced or repaired

Reports of the previously-recalled Bambu Lab 3D printer being a fire hazard resurface, new power board component issue identified as the cause — Bambu says the A1's issue has been addressed, affected units replaced or repaired

La connessione segreta tra microbiota e disturbi psichiatrici

Ricerca genetica identifica una catena causale diretta tra le microorganismi del tratto digestivo e il rischio di sviluppare disturbi psichiatrici gravi. I risultati suggeriscono che specifiche batterie intestinali influenzano lo sviluppo di disturbi come la depressione e l'Alzheimer, alterando livelli di molecole di grasso nel sangue.

Dev uses Claude AI to write a ‘functional NES emulator’ — you can test it now, playing Donkey Kong in your browser

A developer has succeeded in prompting Claude to write 'a functional NES emulator.' Now you can test it playing Donkey Kong in your browser.

Head Transplant Surgery: The Future of Medicine

Sergio Canavero's head transplant surgery idea has been met with skepticism in the past. However, as technology advances, this procedure may become a reality. What does this mean for the future of medicine?

3 cose che stanno facendo Will Douglas Heaven

Le banche europee si preparano a tagliare 200.000 posti di lavoro

La ristrutturazione sarà più intensa nei settori di back-office, gestione del rischio e conformità.

China's president Xi Jinping calls Taiwan reunification 'unstoppable' — military drills around the island escalate in area and deployments

China calls Taiwan reunification 'unstoppable' as military drills proceed

Large study links abdominal obesity to increased risk of migraines

A large longitudinal study conducted in South Korea found that abdominal obesity is a risk factor for the development of migraines in young adults. The analysis suggests that body composition may be a stronger predictor of migraine risk than general weight.

OpenAI bets big on audio as Silicio Valley declares war on screens

OpenAI bets big on audio as Silicio Valley declares war on screens

Come i media di massa vengono manipolati per diffondere false notizie

U.S. allows TSMC to import chipmaking equipment to its China fabs — Samsung, SK hynix likewise receive go signal from Commerce Department

The U.S. Department of Commerce didn't renew the validated end-user status of these chipmakers, requiring them to acquire annual licenses to import chipmaking tools containing U.S. tech into their Chinese fabs.

Ineguaglianza economica: lavoro più lungo, benessere meno

Risultato di un nuovo studio: l'ineguaglianza economica è associata a un aumento dei lavori e una diminuzione del benessere. Ma come si può combattere questo fenomeno?

La psicologia del desiderio di essere vittime: una minaccia per la democrazia

Nuova ricerca suggerisce che i partigiani politici americani che si considerano vittime di ingiustizia sono più propensi a sostegno delle politiche anti-democratiche. L'analisi dei dati ha rivelato un legame tra la percezione della propria gruppo come vittima e il supporto alle politiche anti-democratiche.

ASRock entra nel mercato dei raffreddamenti liquidi AIO con un design senza pompa

ASRock debutterà i suoi AIO a CES 2026, inclusi modelli flagship con schermi LCD.

AI Labor Is Boring. AI Lust Is Big Business

After years of hype about generative AI increasing productivity and making lives easier, 2025 was the year erotic chatbots defined AI’s narrative.

Nuova tecnologia per ottimizzare i LLM

Meta ha lanciato una nuova tecnologia per migliorare i LLM, evitando l'uso di grandi dataset etichettati.

LLM innovano: nuove regole per l'incertezza

Generalized Regularized Evidential Deep Learning Models: Theory and Comprehensive Evaluation

New framework for analyzing the consistency-accuracy relation of LLMs under controlled input variations

A new framework has been introduced for evaluating the consistency-accuracy relation of LLMs under controlled input variations, using multiple-choice benchmarks as a case study. The framework proposes a global metric that combines the CAR curve to quantify the trade-off between accuracy and consistency.

PharmaShip: An Entity-Centric, Reading-Order-Supervised Benchmark for Chinese Pharmaceutical Shipping Documents

A new dataset for the pharmaceutical industry has been launched to test artificial intelligence models. PharmaShip covers three tasks: entity recognition, relation extraction and reading order prediction.

New breakthrough for LLMs: a framework for structured output reliability

Large Language Models (LLMs) are increasingly deployed for structured data generation, yet output consistency remains critical for production applications. A new framework has been introduced to evaluate and improve consistency in LLM-generated structured outputs.

Revolution in Automatic Learning Models

A new automatic learning model, the Coordinate Matrix Machine (CM$^2$), has been presented. This model is designed to improve human intelligence by learning document structures and classifying documents. CM$^2$ offers a Green AI sustainable and optimized solution for CPU environments.

Nuova tecnologia per analizzare le dinamiche sociali

Un nuovo studio propone un framework di apprendimento automatico che può analizzare le dinamiche sociali senza l'uso di dati esterni. HINTS, acronimo di Human Insights Through Networked Time Series, è un modello che extrae fattori umani dai residui delle serie temporali, migliorando la precisione della previsione.

PyBangla at BLP-2025 Task 2: Enhancing Bangla-to-Python Code Generation with Iterative Self-Correction and Multilingual Agents

A team of researchers has developed a new framework for generating code in Bengali using multilingual agents and iterative self-correction. The project, called PyBanglaCodeAct, was presented at the BLP-2025 conference.