Science – Page 35

Mapping the Ride: How Graph Networks Combat Fraud

01.01.2026 by qfx

A new wave of fraud detection techniques leveraging graph neural networks is emerging to protect ride-hailing platforms and their users.

Beyond the Silos: AI Networks for Smarter Financial Crime Detection

01.01.2026 by qfx

Traditional machine learning and deep learning methods typically necessitate separate training on each sub-dataset - a practice mirrored by all baselines except for two federated approaches, which suggests a potential for more generalized learning through shared parameters.

A new framework leverages federated learning and graph analysis to connect fragmented data and dramatically improve anti-money laundering efforts.

Collective Intelligence: Can AI Groups Outsmart Individual Models?

01.01.2026 by qfx

The study demonstrates that forecast accuracy, when stratified by scenario, diverges between independent and deliberative approaches, suggesting a systematic difference in how these methods respond to varying predictive challenges.

New research demonstrates that structured discussion among diverse artificial intelligence systems can significantly improve the accuracy of predictions.

Beyond the Limits of Prediction: Correcting Machine Learning’s Blind Spot

01.01.2026 by qfx

A novel neural network architecture, LatentNN, mitigates attenuation bias in predictions-demonstrated at a signal-to-noise ratio of one-by learning to shift latent values from noisy observations toward true functions [latex] f(x) = 2x [/latex], as evidenced by a decreasing prediction loss coupled with an increasing likelihood of the latent variable itself.

A new method tackles the systematic underestimation of extreme values in machine learning models, improving accuracy in data-rich but uncertain fields.

Mapping the Adolescent Brain to Predict Tobacco Use

01.01.2026 by qfx

The proposed Graph Neural Network Transformer (GNN-TF) architecture integrates fMRI imaging and structured data for classification tasks, employing a “cls” token as a prompt in all transformer models except GPT2-where, adhering to OpenAI’s guidelines, it is appended to the sequence alongside projected sex and age features-to facilitate comprehensive data analysis.

A new machine learning model leverages brain connectivity and personal data to forecast the likelihood of future tobacco use in adolescents.

Decoding Bitcoin: A New Neural Network for Price Forecasting

31.12.2025 by qfx

The proposed PGRU architecture leverages two parallel GRU networks-one for price features and another for structural features-whose fused outputs, processed by a feedforward network, ultimately generate price predictions.

Researchers have developed a parallel gated recurrent unit (GRU) architecture that offers improved accuracy and efficiency in predicting Bitcoin prices.

Following the Money: AI-Powered Paths to Smarter Venture Capital

31.12.2025 by qfx

The path selector operates by systematically evaluating the graph to identify and retrieve the optimal trajectory, effectively navigating a complex network to pinpoint the most efficient route.

New research demonstrates how combining graph-based knowledge with large language models can significantly improve the accuracy of venture capital investment predictions.

When AI Tries to Persuade: Unmasking Manipulative Language Models

31.12.2025 by qfx

The DarkPatterns-LLM corpus comprises 401 entries distributed across seven categories of harmful design patterns, with each category representing a proportional range of 12.0% to 17.2% of the total dataset.

Researchers have developed a new benchmark to assess how readily large language models exhibit manipulative behaviors, going beyond basic safety to reveal the subtle ways they can influence users.

The AI Mirror Test: Can Language Models Know What They’ve Written?

31.12.2025 by qfx

New research reveals that current AI text detectors are easily tricked, raising serious questions about their reliability in educational settings.

Spotting the Unexpected: AI Learns to Detect Rare Driving Risks

31.12.2025 by qfx

The system integrates machine learning and rule-based detection methods to identify anomalous driving scenarios, aiming for effective responses despite the inevitable challenges of real-world deployment and the eventual accumulation of technical debt inherent in any complex framework.

A new approach uses unsupervised learning to identify unusual driving patterns that could signal potential safety hazards.