ProtAgents: protein discovery via large language model multi-agent collaborations combining physics and machine learning

Digital Discovery, Journal Year: 2024, Volume and Issue: 3(7), P. 1389 - 1409

Published: Jan. 1, 2024

ProtAgents is a de novo protein design platform based on multimodal LLMs, where distinct AI agents with expertise in knowledge retrieval, structure analysis, physics-based simulations, and results analysis tackle tasks dynamic setting.

Language: Английский

---

Automating alloy design and discovery with physics-aware multimodal multiagent AI

Proceedings of the National Academy of Sciences, Journal Year: 2025, Volume and Issue: 122(4)

Published: Jan. 24, 2025

The design of new alloys is a multiscale problem that requires holistic approach involves retrieving relevant knowledge, applying advanced computational methods, conducting experimental validations, and analyzing the results, process typically slow reserved for human experts. Machine learning can help accelerate this process, instance, through use deep surrogate models connect structural chemical features to material properties, or vice versa. However, existing data-driven often target specific objectives, offering limited flexibility integrate out-of-domain knowledge cannot adapt new, unforeseen challenges. Here, we overcome these limitations by leveraging distinct capabilities multiple AI agents collaborate autonomously within dynamic environment solve complex materials tasks. proposed physics-aware generative platform, AtomAgents, synergizes intelligence large language (LLMs) collaboration among with expertise in various domains, including retrieval, multimodal data integration, physics-based simulations, comprehensive results analysis across modalities. concerted effort multiagent system allows addressing problems, as demonstrated examples include designing metallic enhanced properties compared their pure counterparts. Our enable accurate prediction key characteristics highlight crucial role solid solution alloying steer development alloys. framework enhances efficiency multiobjective tasks opens avenues fields such biomedical engineering, renewable energy, environmental sustainability.

Language: Английский

---

OpenFOAMGPT: A retrieval-augmented large language model (LLM) agent for OpenFOAM-based computational fluid dynamics

Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(3)

Published: March 1, 2025

This work presents a large language model (LLM)-based agent OpenFOAMGPT tailored for OpenFOAM-centric computational fluid dynamics (CFD) simulations, leveraging two foundation models from OpenAI: the GPT-4o (GPT means Generative Pre-trained Transformer) and chain-of-thought–enabled o1 preview model. Both agents demonstrate success across multiple tasks. While price of token with is six times as that GPT-4o, it consistently exhibits superior performance in handling complex tasks, zero-shot/few-shot case setup to boundary condition modifications, zero-shot turbulence adjustments, code translation. Through an iterative correction loop, efficiently addressed single-phase multiphase flow, heat transfer, Reynolds-averaged Navier–Stokes modeling, eddy simulation, other engineering scenarios, often converging limited number iterations at low costs. To embed domain-specific knowledge, we employed retrieval-augmented generation pipeline, demonstrating how preexisting simulation setups can further specialize subdomains such energy aerospace. Despite great agent, human oversight remains crucial ensuring accuracy adapting shifting contexts. Fluctuations over time suggest need monitoring mission-critical applications. Although our demonstrations focus on OpenFOAM, adaptable nature this framework opens door developing LLM-driven into wide range solvers codes. By streamlining CFD approach has potential accelerate both fundamental research industrial advancements.

Language: Английский

---

Cephalo: Multi‐Modal Vision‐Language Models for Bio‐Inspired Materials Analysis and Design

Advanced Functional Materials, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 5, 2024

Abstract Cephalo is presented as a series of multimodal vision large language models (V‐LLMs) designed for materials science applications, integrating visual and linguistic data enhanced understanding. A key innovation its advanced dataset generation method. trained on integrated image text from thousands scientific papers science‐focused Wikipedia demonstrates it can interpret complex scenes, generate precise descriptions, answer queries about images effectively. The combination encoder with an autoregressive transformer supports natural understanding, which be coupled other generative methods to create image‐to‐text‐to‐3D pipeline. To develop more capable smaller ones, both mixture‐of‐expert model merging are reported. examined in diverse use cases that incorporate biological materials, fracture engineering analysis, protein biophysics, bio‐inspired design based insect behavior. Generative applications include designs, including pollen‐inspired architected well the synthesis material microstructures photograph solar eclipse. Additional fine‐tuning molecular dynamics results demonstrate Cephalo's capabilities accurately predict statistical features stress atomic energy distributions, crack damage materials.

Language: Английский

---

Can ChatGPT Implement Finite Element Models for Geotechnical Engineering Applications?

International Journal for Numerical and Analytical Methods in Geomechanics, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 6, 2025

ABSTRACT This study assesses the capability of ChatGPT to generate finite element code for geotechnical engineering applications from a set prompts. We tested three different initial boundary value problems using hydro‐mechanically coupled formulation unsaturated soils, including dissipation excess pore water pressure through fluid mass diffusion in one‐dimensional space, time‐dependent differential settlement strip footing, and gravity‐driven seepage. For each case, prompting involved providing with necessary information implementation, such as balance constitutive equations, problem geometry, conditions, material properties, spatiotemporal discretization solution strategies. Any errors unexpected results were further addressed prompt augmentation processes until ChatGPT‐generated passed verification/validation test. Our demonstrate that required minimal revisions when FEniCS library, owing its high‐level interfaces enable efficient programming. In contrast, MATLAB generated by necessitated extensive augmentations and/or direct human intervention, it involves significant amount low‐level programming analysis, constructing shape functions or assembling global matrices. Given this task requires an understanding mathematical numerical techniques, suggests while large language model may not yet replace programmers, can greatly assist implementation models.

Language: Английский

---

Facilitating or Inhibiting: A Study on the Impact of Artificial Intelligence on Corporate Greenwashing

Sustainability, Journal Year: 2025, Volume and Issue: 17(5), P. 2154 - 2154

Published: March 2, 2025

As a significant driving force behind the latest wave of technological innovation, artificial intelligence profoundly influences corporate greenwashing while advancing digital and intelligent transformation enterprises. This paper empirically examines impact AI technology on its mechanisms action using text analysis word frequency statistics. study considers references to in annual reports enterprises ESG scores these as samples. The research findings indicate that application can effectively curb occurrence behavior. influence suggest green innovation plays partial mediating role relationship between greenwashing, imitation pressure financial enhance inhibitory effect this Further reveals is particularly pronounced non-state-owned enterprises, large-scale within high-pollution industries. not only enhances existing literature how promote enterprise greening but also offers valuable insights into governments mitigate

Language: Английский

---

AlloyGPT: End-to-end prediction and design of additively manufacturable alloys using an autoregressive language model

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: March 7, 2025

Rapid progress in additive manufacturing of alloys opens opportunities controlling compositions and microstructures at voxel-size resolution complex geometries, thus unlocking unprecedented design performance various critical engineering applications. However, to fully exploit such potential, capable yet efficient models for navigating the vast spaces alloy compositions, structures properties are great research interest. Here, we present AlloyGPT, an autoregressive alloy-specific language model, that learns composition-structure-property relationship generates novel designs additively manufacturable alloys. Specifically, develop grammar convert physics-rich datasets into readable text records both forward prediction inverse tasks. Then, construct a customized tokenizer generative pre-trained transformer (GPT) model master this through training. At deployment, our can accurately predict multiple phase based on given achieving R2 values ranging from 0.86 0.99 test set. When tested beyond learned composition domain, only degrades gradually stable manner. Given desired structures, same suggest meet goals. And balance between diversity accuracy be further tuned stably. Our AlloyGPT presents way integrating comprehensive knowledge terms simultaneously solve tasks with accuracy, robustness. This fundamental will open new avenues accelerate integration material pure or gradient structural manufactured by traditional manufacturing.

Language: Английский

---

Roadmap on data-centric materials science

Modelling and Simulation in Materials Science and Engineering, Journal Year: 2024, Volume and Issue: 32(6), P. 063301 - 063301

Published: May 17, 2024

Abstract Science is and always has been based on data, but the terms ‘data-centric’ ‘4th paradigm’ of materials research indicate a radical change in how information retrieved, handled performed. It signifies transformative shift towards managing vast data collections, digital repositories, innovative analytics methods. The integration artificial intelligence its subset machine learning, become pivotal addressing all these challenges. This Roadmap Data-Centric Materials explores fundamental concepts methodologies, illustrating diverse applications electronic-structure theory, soft matter microstructure research, experimental techniques like photoemission, atom probe tomography, electron microscopy. While roadmap delves into specific areas within broad interdisciplinary field science, provided examples elucidate key applicable to wider range topics. discussed instances offer insights multifaceted challenges encountered contemporary research.

Language: Английский

---

Multimodal Transformer for Property Prediction in Polymers

ACS Applied Materials & Interfaces, Journal Year: 2024, Volume and Issue: 16(13), P. 16853 - 16860

Published: March 19, 2024

In this work, we designed a multimodal transformer that combines both the Simplified Molecular Input Line Entry System (SMILES) and molecular graph representations to enhance prediction of polymer properties. Three models with different embeddings (SMILES, SMILES + monomer, dimer) were employed assess performance incorporating features into architectures. Fine-tuning results across five properties (i.e., density, glass-transition temperature (Tg), melting (Tm), volume resistivity, conductivity) demonstrated dimer configuration as inputs outperformed using only all Furthermore, our model facilitates in-depth analysis by examining attention scores, providing deeper insights relationship between deep learning attributes. We believe shedding light on potential transformers in predicting properties, paves new direction for understanding refining

Language: Английский

---

A MISLEADING GALLERY OF FLUID MOTION BY GENERATIVE ARTIFICIAL INTELLIGENCE

Journal of Machine Learning for Modeling and Computing, Journal Year: 2024, Volume and Issue: 5(2), P. 113 - 144

Published: Jan. 1, 2024

In this technical report, we extensively investigate the accuracy of outputs from well-known generative artificial intelligence (AI) applications in response to prompts describing common fluid motion phenomena familiar mechanics community. We examine a range applications, including Midjourney, Dall·E, Runway ML, Microsoft Designer, Gemini, Meta AI, and Leonardo introduced by prominent companies such as Google, OpenAI, Meta, Microsoft. Our text for generating images or videos include examples "Von Karman vortex street," "flow past an airfoil," "Kelvin-Helmholtz instability," "shock waves on sharp-nosed supersonic body," etc. compare generated these with real laboratory experiments numerical software. findings indicate that AI models are not adequately trained dynamics imagery, leading potentially misleading outputs. Beyond text-to-image/video generation, further explore transition image/video generation using tools, aiming their descriptions phenomena. This report serves cautionary note educators academic institutions, highlighting potential tools mislead students. It also aims inform researchers at renowned companies, encouraging them address issue. conjecture primary reason shortcoming is limited access copyright-protected scientific journals.

Language: Английский

---