Search papers, labs, and topics across Lattice.
The authors introduce ElementsClaw, an agentic framework that integrates Large Atomic Models (LAMs) and Large Language Models (LLMs) to accelerate materials discovery. ElementsClaw orchestrates LAM tools finetuned from a 1B parameter model for atomic-scale computation, guided by LLMs for high-level reasoning, enabling human-interactive discovery workflows. Applied to superconductors, ElementsClaw screened 2.4 million crystals, identified 68,000 high-confidence candidates, and experimentally verified four novel superconductors, demonstrating a significant expansion of the known superconducting space.
AI can now autonomously discover new materials, as demonstrated by the identification and experimental verification of four novel superconductors using an agentic framework that orchestrates large atomic and language models.
The discovery of novel materials is critical for global energy and quantum technology transitions. While deep learning has fundamentally reshaped this landscape, existing predictive or generative models typically operate in isolation, lacking the autonomous orchestration required to execute the full discovery process. Here we present ElementsClaw, an agentic framework for materials discovery that synergizes Large Atomic Models (LAMs) with Large Language Models (LLMs). In response to varied human queries, ElementsClaw orchestrates a suite of LAM tools finetuned from our proposed 1-billion-parameter model Elements for atomic-scale numerical computation, while leveraging LLMs for high-level semantic reasoning. This shift moves AI-driven materials science from isolated processes toward integrated and human interactive discovery. Applied to superconductors, ElementsClaw screens 2.4 million crystals in just 28 GPU hours to identify 68,000 high-confidence candidates (The complete dataset of screened superconductors is available at https://developer.damo-academy.com/material), expanding known superconducting space by orders of magnitude compared to datasets curated over decades. Critically, ElementsClaw achieves a high success rate in identifying superconductors hidden in literature and discovers four novel experimentally verified superconductors, exemplified by Zr3ScRe8 with a transition temperature of 6.8 K and HfZrRe4 at 6.7 K. Together, our results establish a knowledge integrated, autonomously orchestrated, and experimentally grounded paradigm for materials discovery.
DAMO
Tsinghua AI