arXiv:2603.01311v2 Announce Type: replace 
Abstract: The discovery of novel catalysts tailored for particular applications is a major challenge for the twenty-first century. Traditional methods for this include time-consuming and expensive experimental trial-and-error approaches in labs based on chemical theory or heavily computational first-principles approaches based on density functional theory. Recent studies show that deep learning models like graph neural networks (GNNs) can significantly speed up the screening of catalyst materials by many orders of magnitude, with very high accuracy and fidelity. In this work, we introduce Catalyst-Agent, a Model Context Protocol (MCP) server-based, LLM-powered AI agent. It can explore vast material databases using the OPTIMADE API, make structural modifications, calculate adsorption energies using Meta FAIRchem's UMA (GNN) model via FAIRchem's AdsorbML workflow and slab construction, and make useful material suggestions to the researcher in a closed-loop manner, including structural modifications to refine near-miss candidates. It is tested on three pivotal reactions: the oxygen reduction reaction (ORR), the nitrogen reduction reaction (NRR), and the CO2 reduction reaction (CO2RR). Catalyst-Agent achieves a success rate of 33-41% among all the materials it chooses and evaluates, and manages to converge in 1-4 trials per successful material on average. This work demonstrates the potential of AI agents to exercise their planning capabilities and tool use for autonomous catalyst screening workflows.

يمثل تقديم Catalyst-Agent تقدمًا كبيرًا في مجال اكتشاف المحفزات، حيث يستخدم نموذج لغة كبير (LLM) لفحص المحفزات غير المتجانسة بشكل مستقل. يستفيد هذا الوكيل الذكي من واجهة برمجة التطبيقات OPTIMADE لاستكشاف قواعد بيانات المواد الواسعة، مما يعزز كفاءة ودقة فحص المواد المحفزة مقارنة بالطرق التجريبية التقليدية.

La introducción de Catalyst-Agent marca un avance significativo en el campo del descubrimiento de catalizadores, utilizando un modelo de lenguaje grande (LLM) para realizar un cribado autónomo de catalizadores heterogéneos. Este agente de IA aprovecha la API OPTIMADE para explorar extensas bases de datos de materiales, mejorando la eficiencia y precisión del cribado de materiales catalíticos en comparación con los métodos experimentales tradicionales.

L'introduction de Catalyst-Agent représente une avancée significative dans le domaine de la découverte de catalyseurs, utilisant un modèle de langage large (LLM) pour dépister de manière autonome des catalyseurs hétérogènes. Cet agent IA exploite l'API OPTIMADE pour explorer d'immenses bases de données de matériaux, améliorant l'efficacité et la précision du dépistage des matériaux catalytiques par rapport aux méthodes expérimentales traditionnelles.

The introduction of Catalyst-Agent marks a significant advancement in the field of catalyst discovery, utilizing a large language model (LLM) to autonomously screen heterogeneous catalysts. This AI agent leverages the OPTIMADE API to explore extensive material databases, enhancing the efficiency and accuracy of catalyst material screening compared to traditional experimental methods.

Catalyst-Agent: Autonomous heterogeneous catalyst screening with an LLM Agent

arXiv:2605.13268v2 Announce Type: replace-cross 
Abstract: Trotter Suzuki product formulas are the standard route to Hamiltonian evolution on noisy intermediate-scale quantum (\NISQ{}) hardware, but their accuracy depends on three coupled choices: term grouping, product-formula order, and time-step allocation. Grouping and order are discrete, which makes direct gradient optimization infeasible and forces existing compilers to rely on static heuristics.
  We describe P-GONE, a method that combines a conditional diffusion model (D3PM + DDPM), a graph neural network (\GNN{}) encoder, and closed-loop REINFORCE fine-tuning to jointly learn grouping, order, and time-step optimization over a mixed discrete-continuous space. Under fidelity-matched conditions ($F \geq 0.95$), the method achieves circuit depth 86 versus 1673 for Qiskit fourth-order (ungrouped, Suzuki-4), about $19.4\times$ compression, and 141 for Paulihedral (first-order Trotter), about $1.6\times$ compression. At $T=0.90$ the method also beats the Qiskit group-commuting teacher (65 vs 103, $1.6\times$ compression), though at $T=0.95$ the teacher still leads -- a stratified pattern that points toward fidelity-aware fine-tuning. Under a standard depolarizing noise model, the method achieves noisy fidelity roughly $2\times$ the Qiskit fourth-order baseline (0.743 vs 0.380). Ablation shows a clear hierarchy: order learning $>$ time allocation $>$ grouping. Best-of-N sampling ($N=32$ is a practical sweet spot) and CFG guidance give flexible fidelity-depth trade-offs at inference. The method works well on structured Hamiltonians (TFIM, Heisenberg), but random Pauli Hamiltonians fail entirely at $T \geq 0.95$ -- a boundary that defines where the method applies.

تم تقديم طريقة جديدة تسمى P-GONE لتحسين تحليل Trotter Suzuki لتطور هاملتونيان على الأجهزة الكمومية ذات النطاق المتوسط المليئة بالضوضاء (NISQ). تجمع هذه الطريقة بين نموذج انتشار شرطي، ومشفر شبكة عصبية رسومية، وتعلم تعزيز مغلق لتحسين دقة تصميم الدوائر الكمومية من خلال تحسين تجميع المصطلحات، وترتيب صيغة المنتج، وتخصيص خطوات الزمن.

Se ha introducido un nuevo método llamado P-GONE para optimizar la descomposición de Trotter Suzuki para la evolución Hamiltoniana en hardware cuántico de escala intermedia ruidoso (NISQ). Este método combina un modelo de difusión condicional, un codificador de red neuronal gráfica y un aprendizaje por refuerzo en bucle cerrado para mejorar la precisión del diseño de circuitos cuánticos al optimizar el agrupamiento de términos, el orden de la fórmula del producto y la asignación de pasos de tiempo.

Une nouvelle méthode appelée P-GONE a été introduite pour optimiser la décomposition de Trotter Suzuki pour l'évolution Hamiltonienne sur du matériel quantique à échelle intermédiaire bruité (NISQ). Cette méthode combine un modèle de diffusion conditionnel, un encodeur de réseau de neurones graphiques et un apprentissage par renforcement en boucle fermée pour améliorer l'exactitude de la conception des circuits quantiques en optimisant le regroupement des termes, l'ordre des formules de produit et l'allocation des pas de temps.

A new method called P-GONE has been introduced to optimize Trotter Suzuki decomposition for Hamiltonian evolution on noisy intermediate-scale quantum (NISQ) hardware. This method combines a conditional diffusion model, a graph neural network encoder, and closed-loop reinforcement learning to enhance the accuracy of quantum circuit design by optimizing term grouping, product-formula order, and time-step allocation.

Catalyst-Agent: Autonomous heterogeneous catalyst screening with an LLM Agent

Was this article worth reading? Share it

Ready to build your own newsroom?