Tag
The paper proposes a data-driven constitutive modeling framework using neural operators with causal attention to predict stress response in path-dependent materials, focusing on plasticity and damage, with accurate and parallelizable predictions.
A new study finds that charged raindrops can electrically damage protective coatings like Teflon, leading to corrosion through a spark mechanism similar to lightning bolts.
Gemini Co-Scientist, a Gemini-based multi-agent system, has been validated in real-world scientific research across materials science, biology, and computer science, demonstrating capabilities in experimental design, outcome prediction, and improving AI-generated research.
Packora is a flow-based generative model for molecular crystal structure prediction that jointly predicts atomic coordinates and lattice from molecular graphs, outperforming baselines in generation and ranking benchmarks.
Research published in Nature reveals that water droplets become electrified when sliding over smooth surfaces, leading to corrosion in metals.
SchemaRouter introduces a lightweight schema-graph routing layer for heterogeneous agentic RAG systems, minimizing retrieval overhead by selecting only necessary tools and fields while maintaining answer accuracy and provenance attribution.
MIT researchers have developed a predictive approach to identify promising catalysts for electrochemical ammonia production, aiming to make this sustainable method competitive with the energy-intensive Haber-Bosch process.
Scientists used advanced X-ray imaging techniques to discover a double-helix structure in narwhal tusks, which contributes to their strength and may record environmental changes over time.
This paper proposes a self-evolving memory framework for lifelong AI partners in materials science, storing scientific experience as facts and skills to improve agent performance across models. Evaluations show memory nearly doubles GPT-5.2 task success in materials tool-use questions and reduces repeated errors in simulations.
This paper reports on the study of superconductivity in a monolayer cuprate material with a single CuO2 plane, which could advance understanding of high-temperature superconductors.
This paper introduces DyRIS, an LLM-agent framework using dynamic few-shot retrieval and rule-guided inference to predict space groups of double perovskites, achieving competitive accuracy and significantly improving performance on minority space-group classes.
Discovered Materials, a YC-backed startup, uses AI agents and physics models to discover new semiconductor materials that could run cooler, aiming to reduce data center power consumption. It closed a $9M seed round and released a benchmark called Material Discovery Bench.
Introduces CrystalGRPO, a reinforcement-learning post-training framework for flow-based crystal structure prediction that aligns target recovery and preserves candidate coverage, improving Top-1 and Top-20 performance across MP-20 and MPTS-52 benchmarks.
ED-CSP is a machine learning model that predicts crystal structures from electron diffraction patterns, achieving improved match rates over the PXRD-based PXRDGen and demonstrating the value of multi-view diffraction data.
This paper introduces SEE, a multimodal benchmark of expert-curated questions for scientific discovery in chemistry, biology, and materials science. Evaluation of 19 MLLMs shows the best model reaches only 48.7% accuracy, and even with tool use only 52.7%, revealing that current models lack reliable evidence-bounded scientific reasoning.
This paper introduces CGMas, a multi-agent LLM framework that automates coarse-grained molecular dynamics for polymers, including topology construction, equilibration, mapping, potential derivation, and validation. It completed 27 polymer tasks and matched atomistic densities within 5% in most cases, drastically reducing simulation time.
Researchers analyzed five types of snail mucus, finding that collagen IV forms the structural base while protein content and amorphous calcium carbonate tune each mucus's mechanical properties, including defensive sticky and bubbly secretions.
This paper presents a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a materials-science hypothesis generation model, using visualization and activation-based diagnostics to localize where mechanism support is lost or recovered during generation.
This paper proposes using atom-averaged features from pretrained MLIPs like MACE as coarse coordinates for evaluating and guiding inorganic crystal structure generation, introducing the Coarse-Fine Transport Distance (CFTD) metric that captures both quality and novelty in a distribution-based framework.
A new multicomponent alloy forged by the Hiroshima atomic blast has been discovered, providing unique insights into material behavior under extreme conditions.