Science – Page 43

Can AI Pilot Complex Simulations?

15.02.2026 by ebaster

An autonomous agent leverages function calls to external computational tools-specifically OpenFOAM, Gmsh, and Python-within an iterative workflow guided by a specialized prompt prioritizing tutorial reuse and log-driven repair, effectively navigating complex simulations until a defined end time is achieved, demonstrating a system capable of self-correction and persistent execution.

This research investigates automating computational fluid dynamics workflows with artificial intelligence, aiming to improve the reliability of complex engineering simulations.

Grok’s Social Life: How AI is Finding Its Role Online

15.02.2026 by ebaster

Grok’s ten defined roles-plotted across axes of allocentric versus egocentric perspectives and supportive versus constraining behaviors-demonstrate an emergence on social media, with roles exhibiting both characteristics highlighted in yellow.

A new study examines how users are interacting with the Grok large language model on X, revealing the distinct social roles the AI is beginning to assume.

Decoding the Language of Life: Mapping Protein Circuits with AI

15.02.2026 by ebaster

ProtoMech introduces cross-layer transcoders that predict outputs based on sparse latent features, enabling the design of protein variants with improved function by steering along identified biological circuits and revealing previously hidden motifs within ESM2 models.

Researchers have developed a new framework to dissect the inner workings of protein language models, revealing the computational circuits that drive biological function.

Reasoning Machines: Unleashing Algorithmic Power in Large Language Models

14.02.2026 by ebaster

The PRIME framework establishes a system where reasoning steps are continuously vetted for consistency, with a coordinating mechanism managing iterative refinement through a state-based backtracking process informed by Group Relative Policy Optimization [latex] GRPO [/latex], acknowledging that even robust systems require ongoing recalibration to maintain integrity over time.

A new framework combines the strengths of multi-agent systems and iterative refinement to dramatically improve the ability of AI to solve complex problems.

Unlocking Plasma Physics Knowledge with AI

14.02.2026 by ebaster

The annotated dataset reveals a nuanced approach to entity recognition, where terms are intentionally layered with multiple, overlapping classifications to reflect the inherent complexity of information.

A new approach to information extraction leverages artificial intelligence to identify and categorize key concepts within complex research articles.

Charting a Faster Course: Multi-Graph Search for Robot Navigation

14.02.2026 by ebaster

The system generates navigable environments represented as occupancy grids-used to compute attractive forces during path planning-and successfully demonstrates its capabilities beyond robotic manipulation by autonomously charting a course through a complex three-dimensional space.

A new motion planning framework dramatically improves efficiency by simultaneously exploring multiple potential paths for robots operating in complex, high-dimensional spaces.

Building Robots That Understand the World: LDA-1B Learns From Experience

14.02.2026 by ebaster

Latent Dirichlet Allocation (LDA) demonstrates scalable performance in action prediction, achieving 6.6 error with 30,000 hours of training-a result attributable to effective data diversification-and consistently surpasses the Unsupervised Weighting Model (UWM) across parameter scales-from 0.1 to 1 billion-while the baseline method rapidly plateaus in its learning capacity.

A new foundation model, LDA-1B, is pushing the boundaries of embodied AI by learning robust interaction dynamics from a vast and diverse range of real-world robot data.

Beyond Fact-Checking: How Humans and AI Can Reason Together

14.02.2026 by ebaster

A new system called Althea demonstrates that collaborative exploration, rather than simple AI assistance, is key to improving critical thinking and building trust in information.

Building Smarter AI Agents: A System for Adaptive Configuration

14.02.2026 by ebaster

The method navigates a vast configuration space-spanning thousands of possibilities-to optimize performance for the Qwen 2.5 7B Instruct model across multiple datasets, demonstrating a capacity to adapt and improve beyond baseline results.

New research details a framework for automatically optimizing the workflows and instructions of AI agents powered by large language models.

The AI Negotiation Edge: Control vs. Collective Gain

14.02.2026 by ebaster

Delegate agents demonstrably improve trade outcomes by consistently proposing offers that yield a significant increase in receiver surplus-a statistically significant shift not observed in non-AI-assisted scenarios [latex] (p<0.01) [/latex], suggesting a capacity for mutually beneficial negotiation.

New research explores how different levels of AI involvement in multi-party bargaining – from advice to full delegation – impacts strategic choices and overall outcomes.