Science – Page 33

Beyond Black Boxes: Bridging AI and Statistics for Smarter Models

02.01.2026 by qfx

Despite utilizing fewer features, a linear model demonstrably outperforms a neural network in cold-start recommendation on the Netflix dataset-achieving a cosine similarity of approximately 0.59 compared to 0.52-suggesting the relationship between semantic features and collaborative filtering embeddings is largely linear, and exceeding the performance of a GPT-5 zero-shot baseline (CS=0.5).

A new approach combines the predictive power of large language models with the interpretability of traditional statistical methods to unlock deeper insights from complex data.

Reading the Game: How AI Develops Intuition in Poker

02.01.2026 by qfx

Dimensionality reduction via UMAP reveals semantic clustering of hand-rank activations across the initial transformer layers, though this representation introduces distortion as training set size varies, suggesting a trade-off between interpretability and fidelity in the learned feature space.

New research shows that artificial intelligence trained on the complex game of poker can build internal models of the game state, including probabilistic beliefs about hidden information.

Can AI Truly Predict the Future, or Just Recall the Past?

02.01.2026 by qfx

Pairs Bootstrap Inference reveals the distribution of effects estimated from observational data, allowing for robust assessment of statistical significance without reliance on strong parametric assumptions.

New research reveals that a surprising amount of apparent forecasting ability in large language models may stem from memorization rather than genuine predictive power.

Decoding Bias in News: How AI Explanations Differ

02.01.2026 by qfx

The bias-detector model exhibits predictable sensitivities, as indicated by specific top bias indicators.

A new study compares how two artificial intelligence models identify bias in news articles, revealing crucial differences in their reasoning.

Power to the Prosumers: Decentralized Energy Markets Emerge

02.01.2026 by qfx

The interplay between a power grid and its constituent prosumers reveals a dynamic of reciprocal exchange, where aggregated energy sales to the grid-indicated by increasing green values-are balanced against community energy purchases-represented by growing red values-with the resulting net interaction-traced by the blue line-highlighting periods of energy surplus and deficit.

A new approach to energy sharing leverages automated market making to efficiently coordinate distributed resources and empower individual producers.

Simulating Particle Physics with AI: A New Approach to Detector Modeling

02.01.2026 by qfx

The detector system maps the trajectory of a muon-a fleeting particle born of cosmic rays-as it traverses two distinct layers of silicon, each sensitive to the particle’s passage, revealing subtle deflections indicative of interactions within the material and providing a precise record of its path through matter.

Researchers are exploring the use of deep learning models, inspired by the architecture behind large language models, to accelerate the computationally intensive process of simulating silicon tracking detectors.

When Machine Learning Meets Soil: A Reality Check for Geotechnical Engineering

02.01.2026 by qfx

The Deep Operator Network architecture learns a functional mapping from input loads to structural deflections by encoding both the applied load [latex] p(x) [/latex] and query locations [latex] x [/latex] into latent representations, subsequently combining these via a dot product to rapidly predict deflections [latex] w(x) [/latex] for arbitrary load patterns following offline training.

A new review assesses the current capabilities of deep learning techniques, specifically Physics-Informed Neural Networks and Operator Learning, for solving complex problems in soil mechanics and foundation design.

Balancing Risk and Equity: The Future of Insurance Pricing

02.01.2026 by qfx

A new optimization framework seeks to reconcile accurate risk assessment with critical fairness considerations in insurance coverage.

Quieting the Noise: Aligning Audio AI for Reliable Reasoning

02.01.2026 by qfx

$The system constructs a unified data pipeline for enhancing audio-question answering by establishing a shared dataset [latex]\mathcal{D}[/latex] and then diverging into two views: one aligning responses through preference pairs contrasting chosen and hallucination-containing rejected answers, and another rigorously evaluating performance with human-verified ground truth and annotated hallucination types [latex]\tau(q\_{i})[/latex].$

New research tackles the problem of ‘audio hallucinations’ in large AI models, improving their ability to accurately process and understand spoken information.

Predicting the Sun: AI Improves Solar Power Forecasting

02.01.2026 by qfx

Across Vietnam on March 19, 2018, at 12:00 VNT, the distribution of global horizontal irradiance-a measure of solar radiation-varied significantly by region, demonstrating the complex interplay of geography and atmospheric conditions in capturing solar energy.

New research showcases how advanced artificial intelligence models are significantly enhancing the accuracy of short-term solar irradiance predictions, paving the way for a more stable and efficient renewable energy grid.