Agentic Fusion of Large Atomic and Language Models to Accelerate Superconductor Discovery

• URL: http://arxiv.org/abs/2604.23758v3
• Date: Mon, 04 May 2026 03:23:33 GMT
• Title: Agentic Fusion of Large Atomic and Language Models to Accelerate Superconductor Discovery
• Abstract summary: Scaling to 2.4 million equilibrium crystals, ElementsClaw identifies 68,000 high-confidence candidates in just 28 GPU hours.<n>We experimentally synthesize and verify four novel superconductors.
• Score: 85.54953285509167
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Artificial intelligence has accelerated materials discovery through high-throughput prediction and generation, yet the decision problem remains a formidable bottleneck. While current AI systems readily propose millions of candidates, navigating the decision regarding a viable experimental target requires resolving multi-dimensional judgments across atomic-scale numerical computation and high-level semantic reasoning. Here we present ElementsClaw, an agentic framework for materials discovery that orchestrates a suite of Large Atomic Model (LAM) tools finetuned from our proposed 1-billion-parameter model Elements for numerical computation, while leveraging Large Language Models (LLMs) for semantic reasoning. Applied to superconductors, ElementsClaw rediscovers 66 experimentally verified superconductors that are absent from the standard SuperCon3D database. Scaling to 2.4 million equilibrium crystals, ElementsClaw identifies 68,000 high-confidence candidates in just 28 GPU hours (https://developer.damo-academy.com/material), expanding known superconducting space by orders of magnitude compared to datasets curated over decades. Guided by the agent's reasoning, we experimentally synthesize and verify four novel superconductors: the motif-guided Zr$_3$ScRe$_8$ ($T_c$ = 6.5 K), the de novo generated HfZrRe$_4$ ($T_c$ = 5.9 K), the structurally reinterpreted Zr$_4$VRe$_7$ ($T_c$ = 3.5 K), and the database-latent Hf$_{21}$Re$_{25}$ ($T_c$ = 2.5 K). Together, our results establish a knowledge integrated, autonomously orchestrated, and experimentally grounded paradigm for materials discovery.

Related papers

• Attributing Emergence in Million-Agent Systems [68.53670424791751]
  Large language models (LLMs) can simulate human-like reasoning and decision-making in individual agents.<n>Such studies require attributing macro emergence to individual agents.<n>We adapt Aumann--Shapley path-integral attribution to LLM-powered MAS at million-agent scale.
  arXiv  Detail & Related papers  (2026-05-12T01:49:41Z)
• Guided Diffusion for the Discovery of New Superconductors [4.592996089024724]
  inverse design of materials with specific desired properties, such as high-temperature superconductivity, represents a formidable challenge in materials science.<n>We present a guided diffusion framework to accelerate the discovery of novel superconductors.
  arXiv  Detail & Related papers  (2025-09-29T17:59:52Z)
• TopoMAS: Large Language Model Driven Topological Materials Multiagent System [9.394163894876689]
  TopoMAS is an interactive human-AI framework that seamlessly orchestrates the materials-discovery pipeline.<n>TopoMAS closes the loop by autonomously integrating computational outcomes into a dynamic knowledge graph.<n>It has already guided the identification of novel topological phases SrSbO3.
  arXiv  Detail & Related papers  (2025-07-05T14:23:12Z)
• Solving the Hubbard model with Neural Quantum States [66.55653324211542]
  We study the state-of-the-art results for the doped two-dimensional (2D) Hubbard model.<n>We find different attention heads in the NQS ansatz can directly encode correlations at different scales.<n>Our work establishes NQS as a powerful tool for solving challenging many-fermions systems.
  arXiv  Detail & Related papers  (2025-07-03T14:08:25Z)
• HTSC-2025: A Benchmark Dataset of Ambient-Pressure High-Temperature Superconductors for AI-Driven Critical Temperature Prediction [5.048163814984219]
  We present the HTSC-2025, an ambient-pressure high-temperature superconducting benchmark dataset.<n>This benchmark holds significant importance for accelerating the discovery of superconducting materials using AI-based methods.
  arXiv  Detail & Related papers  (2025-06-04T11:14:00Z)
• Siamese Foundation Models for Crystal Structure Prediction [70.63218101004398]
  Crystal Structure Prediction (CSP) aims to generate stable crystal structures from compositions.<n>CSP remains a relatively underexplored area.<n>We propose Siamese foundation models specifically designed to address CSP.
  arXiv  Detail & Related papers  (2025-03-13T15:44:16Z)
• InvDesFlow: An AI-driven materials inverse design workflow to explore possible high-temperature superconductors [9.926621857444765]
  InvDesFlow is an AI search engine that integrates deep model pre-training and fine-tuning techniques, diffusion models, and physics-based approaches.<n>We have obtained 74 dynamically stable materials with critical temperatures predicted by the AI model to be $T_c geq$ 15 K based on a very small set of samples.
  arXiv  Detail & Related papers  (2024-09-12T14:16:56Z)
• Inertial Confinement Fusion Forecasting via Large Language Models [48.76222320245404]
  In this study, we introduce $textbfLPI-LLM$, a novel integration of Large Language Models (LLMs) with classical reservoir computing paradigms. We propose the $textitLLM-anchored Reservoir$, augmented with a $textitFusion-specific Prompt$, enabling accurate forecasting of $textttLPI$-generated-hot electron dynamics during implosion. We also present $textbfLPI4AI$, the first $textttLPI$ benchmark based
  arXiv  Detail & Related papers  (2024-07-15T05:46:44Z)
• Accelerating superconductor discovery through tempered deep learning of the electron-phonon spectral function [0.0]
  We train a deep learning model to predict the electron-phonon spectral function, $alpha2F(omega)$. We then incorporate domain knowledge of the site-projected phonon density states to impose inductive bias into the model's node attributes and enhance predictions. This methodological innovation decreases the MAE to 0.18, 29 K, and 28 K, respectively yielding an MAE of 2.1 K for $T_c$.
  arXiv  Detail & Related papers  (2024-01-29T22:44:28Z)
• Towards large-scale quantum optimization solvers with few qubits [59.63282173947468]
  We introduce a variational quantum solver for optimizations over $m=mathcalO(nk)$ binary variables using only $n$ qubits, with tunable $k>1$. We analytically prove that the specific qubit-efficient encoding brings in a super-polynomial mitigation of barren plateaus as a built-in feature.
  arXiv  Detail & Related papers  (2024-01-17T18:59:38Z)
• Design and simulation of a transmon qubit chip for Axion detection [103.69390312201169]
  Device based on superconducting qubits has been successfully applied in detecting few-GHz single photons via Quantum Non-Demolition measurement (QND) In this study, we present Qub-IT's status towards the realization of its first superconducting qubit device.
  arXiv  Detail & Related papers  (2023-10-08T17:11:42Z)
• Robust Hamiltonian Engineering for Interacting Qudit Systems [50.591267188664666]
  We develop a formalism for the robust dynamical decoupling and Hamiltonian engineering of strongly interacting qudit systems. We experimentally demonstrate these techniques in a strongly-interacting, disordered ensemble of spin-1 nitrogen-vacancy centers.
  arXiv  Detail & Related papers  (2023-05-16T19:12:41Z)
• Halide perovskite artificial solids as a new platform to simulate collective phenomena in doped Mott insulators [43.55994393060723]
  We introduce artificial lattices made of lead halide perovskite nanocubes as a new platform to simulate and investigate the physics of correlated quantum materials. We show that, at large photo-doping, the exciton gas undergoes an excitonic Mott transition, which fully realizes the magnetic-field-driven insulator-to-metal transition described by the Hubbard model. Our results demonstrate that time-resolved experiments span a parameter region of the Hubbard model in which long-range and phase-coherent orders emerge out of a doped Mott insulating phase.
  arXiv  Detail & Related papers  (2023-03-15T17:38:51Z)
• Quantum Simulation of an Extended Fermi-Hubbard Model Using a 2D Lattice of Dopant-based Quantum Dots [5.046411982403706]
  The Hubbard model is one of the primary models for understanding the essential many-body physics in condensed matter systems. Recent advances in atomically precise fabrication in silicon have made possible atom-by-atom fabrication of single and few-dopant quantum dots. We demonstrate the analog quantum simulation of a 2D extended Fermi-Hubbard Hamiltonian using STM-fabricated 3x3 arrays of single/few-dopant quantum dots.
  arXiv  Detail & Related papers  (2021-10-18T02:24:05Z)
• BIGDML: Towards Exact Machine Learning Force Fields for Materials [55.944221055171276]
  Machine-learning force fields (MLFF) should be accurate, computationally and data efficient, and applicable to molecules, materials, and interfaces thereof. Here, we introduce the Bravais-Inspired Gradient-Domain Machine Learning approach and demonstrate its ability to construct reliable force fields using a training set with just 10-200 atoms.
  arXiv  Detail & Related papers  (2021-06-08T10:14:57Z)
• Exploring Sparse Expert Models and Beyond [51.90860155810848]
  Mixture-of-Experts (MoE) models can achieve promising results with outrageous large amount of parameters but constant computation cost. We propose a simple method called expert prototyping that splits experts into different prototypes and applies $k$ top-$1$ routing. This strategy improves the model quality but maintains constant computational costs, and our further exploration on extremely large-scale models reflects that it is more effective in training larger models.
  arXiv  Detail & Related papers  (2021-05-31T16:12:44Z)
• Interpretable discovery of new semiconductors with machine learning [10.09604500193621]
  We report an evolutionary algorithm powered search which uses machine-learned surrogate models trained on hybrid functional DFT data benchmarked against experimental bandgaps: Deep Adaptive Regressive Weighted Intelligent Network (DARWIN) The strategy enables efficient search over the materials space of 10$8$ ternaries and 10$11$ quaternaries$7$ for candidates with target properties.
  arXiv  Detail & Related papers  (2021-01-12T10:23:16Z)
• Quantum Sensors for Microscopic Tunneling Systems [58.720142291102135]
  tunneling Two-Level-Systems (TLS) are important for micro-fabricated quantum devices such as superconducting qubits. We present a method to characterize individual TLS in virtually arbitrary materials deposited as thin-films. Our approach opens avenues for quantum material spectroscopy to investigate the structure of tunneling defects.
  arXiv  Detail & Related papers  (2020-11-29T09:57:50Z)
• Bayesian Force Fields from Active Learning for Simulation of Inter-Dimensional Transformation of Stanene [3.708456605408296]
  We present a way to dramatically accelerate Gaussian process models for interatomic force fields based on many-body kernels. This allows for automated active learning of models combining near-quantum accuracy, built-in uncertainty, and constant cost of evaluation.
  arXiv  Detail & Related papers  (2020-08-26T20:27:19Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.