Science – Page 80

Warming Up AI: How Abstract Data Boosts Language Model Reasoning

01.02.2026 by ebaster

Language models benefit from an initial training phase on algorithmically generated procedural data-designed to instill fundamental skills-which accelerates subsequent standard pretraining and enhances performance across varied domains, with different architectural layers contributing uniquely to expertise in each.

A new study reveals that pretraining language models on abstract procedural data significantly improves their ability to tackle complex tasks and learn more efficiently.

AI-Powered Classrooms: The Future of Collaborative Learning

01.02.2026 by ebaster

Large language models now contribute to the iterative refinement of systems, generating feedback that anticipates eventual systemic failure rather than simply optimizing for present performance.

A new framework harnesses the power of advanced artificial intelligence to create dynamic and personalized learning experiences for students.

Boosting Tables with Brains and AI

01.02.2026 by ebaster

Across six distinct tasks, a novel feature engineering method demonstrably outperforms both CAAFE and OCTree-two leading large language model baselines-within a 50-iteration budget when paired with a multilayer perceptron, as indicated by consistently lower error margins represented by the shaded standard error.

A new framework combines human intuition with the power of large language models to create better features for tabular data.

AI Designs Better Ship Propellers

01.02.2026 by ebaster

The computational fluid dynamics model generated two distinct propeller designs, demonstrating its capacity for diverse aerodynamic solutions.

New research harnesses the power of generative artificial intelligence to create high-performance ship propellers tailored to specific needs.

Intelligent Networks: Orchestrating AI at the Edge

01.02.2026 by ebaster

The convergence of multi-agent artificial intelligence within the framework of ubiquitous 6G networks is enabling a new paradigm of distributed intelligence.

A new framework leverages the power of collaborative AI agents and next-generation 6G networks to deliver low-latency intelligence for critical applications.

The AI-Powered Developer: A New Era of Code Creation

01.02.2026 by ebaster

A new study examines how generative AI tools are changing the landscape of software development and impacting developer workflows.

Robots That Reason: Predicting the Future for Smarter Control

01.02.2026 by ebaster

$LingBot-VA constructs a predictive world model through pretraining on expansive video and robotic action data, achieving substantial performance gains on complex real-world manipulation tasks-including those involving deformable objects and precision control-and demonstrating emergent capabilities such as long-range temporal memory and few-shot adaptation beyond standard policy learning, effectively modeling visual dynamics and inferring inverse dynamics from robot videos-surpassing benchmarks like [latex]\pi_{0.5}[/latex].$

New research demonstrates a powerful framework for robots to anticipate outcomes and adapt their actions, leading to more robust and generalizable manipulation skills.

How Your First Language Shapes the Second

01.02.2026 by ebaster

The study models command-line interface interactions by strategically introducing a second language at defined points during training, following an initial pretraining phase on a first language, and then interleaving sequences from both to preserve the influence of the original language-a technique applied to five languages selected to represent increasing degrees of syntactic divergence.

New research uses artificial intelligence to reveal the complex interplay between languages in the human mind.

Beyond the Familiar: Injecting Novelty into AI-Generated Imagery

01.02.2026 by ebaster

The system successfully generates creative visual representations of common objects-specifically, buildings and chairs-demonstrating its capacity for imaginative synthesis.

Researchers are exploring methods to guide diffusion models towards generating truly original images, moving beyond predictable outputs.

Can AI Truly Reason with Facts?

01.02.2026 by ebaster

A new benchmark probes the ability of large language models to effectively integrate retrieved knowledge into complex scientific problem-solving.