Science – Page 165

Unfolding ESMFold: How AI Predicts Protein Shapes

07.02.2026 by ebaster

ESMFold leverages a protein language model-specifically ESM-2-to initially encode amino acid sequences, then iteratively refines both sequence and pairwise representations across 4848 processing blocks before translating them into three-dimensional residue coordinates, effectively growing a structural prediction rather than constructing it.

New research sheds light on the inner workings of the ESMFold model, revealing the computational steps behind its remarkable ability to predict protein structures.

Can AI Agents Truly Interact With the World?

06.02.2026 by ebaster

The system cultivates a continuously evolving landscape of tasks, wherein an agent’s interactions with a simulated environment yield data used to both filter and recombine existing challenges, ensuring ongoing diversity through an ε-random walk, and ultimately generating new objectives-such as navigation or visibility checks-from the remnants of prior iterations.

New research reveals a significant gap between AI performance on standard benchmarks and its ability to handle complex, real-world tasks in 3D environments.

Charting the Unlikely: Machine Learning Accelerates Rare Event Simulation

06.02.2026 by ebaster

AIMMD iteratively trains a neural network with transition path sampling to predict a committor function, which is then expressed in terms of the system's physical collective variables via symbolic regression.

A new approach combines path sampling with artificial intelligence to automatically identify key molecular pathways and dramatically speed up the study of infrequent, yet crucial, events.

Rewriting Causality with Physics

06.02.2026 by ebaster

A new approach leverages physical laws to unlock causal relationships hidden within complex, changing data streams.

Flow with the Crowd: Smarter Robot Navigation in Dense Spaces

06.02.2026 by ebaster

The autonomous system demonstrates graceful navigation within a dense pedestrian environment by aligning with prevailing flow-initially tracking forward movement and subsequently adapting to avoid opposing groups-achieving continued, safe progress as evidenced by its trajectory between [latex]t=4[/latex] and [latex]t=17[/latex].

A new framework allows robots to navigate crowded environments by predicting pedestrian movement and aligning with natural flow.

Beyond Privacy Settings: Building AI Agents That Understand Your Boundaries

06.02.2026 by ebaster

A dynamic privacy agent proactively manages a user’s data-sharing history, aligning past disclosures with current preferences-a concept highly valued among human-centered privacy management designs.

New research explores how AI-powered tools can move beyond pre-sharing privacy controls to dynamically manage personal data and respect evolving preferences across digital interactions.

Beyond Trial and Error: Scaling Collective Intelligence with Scientific Rigor

06.02.2026 by ebaster

The research demonstrates a transition from stochastic performance gains-previously achieved through opaque, untraceable methods-to a scientifically guided approach where factor attribution is inherent in a system that filters resource accumulation to isolate genuine collaborative benefits, represented by Γ, a prism-like analytic step.

As AI agents increasingly collaborate, researchers are shifting focus from simply observing emergent behavior to establishing a science for understanding and maximizing the benefits of teamwork.

Robots to the Rescue: Scaling Urban Search with Affordable Fleets

06.02.2026 by ebaster

New research explores how deploying teams of ground robots, enhanced by artificial intelligence, can dramatically improve the effectiveness of urban search and rescue operations.

Simulating Complexity: From Agents to Adaptive Clusters

06.02.2026 by ebaster

The PhysicsAgentABM architecture performs inference at a cluster level, integrating symbolic and neural pathways with uncertainty-aware fusion before realizing stochastic behaviors at the agent level, and leverages ANCHOR to ensure resulting abstractions remain behaviorally coherent.

A new framework combines the power of artificial intelligence with agent-based modeling to create more accurate and scalable simulations of dynamic systems.

Robots That Hand You Things: A Smarter Approach to Collaboration

06.02.2026 by ebaster

Researchers have developed a new framework enabling robots to more effectively and intuitively hand objects to humans, even when faced with unfamiliar tasks or items.