2024年大型语言模型（LLM）材料科学与化学应用黑客松反思
机器学习
2025-01-06 v2 材料科学
化学物理
摘要
本文介绍了第二届大型语言模型（LLM）材料科学与化学应用黑客松的结果，该活动吸引了跨全球混合地点的参与者，产生了34支团队的提交。提交作品涵盖七个关键应用领域，展示了LLM在材料科学与化学研究中的多样化用途：（1）分子和材料属性预测；（2）分子和材料设计；（3）自动化和新颖接口；（4）科学通信和教育；（5）研究数据管理和自动化；（6）假设生成和评估；（7）从科学文献中提取和推理。每支团队的提交均以汇总表格形式呈现，附有代码链接和附录中的简要论文。除了团队结果外，我们还讨论了黑客松活动及其混合格式，包括在多伦多、蒙特利尔、旧金山、柏林、洛桑和东京的物理中心，以及全球在线中心，以促进本地和虚拟协作。总体而言，活动凸显了自去年举办的黑客松以来LLM能力的显著提升，表明LLM为材料科学和化学研究的应用将继续扩张。这些结果表明，LLM既是多功能模型，可用于多样化的机器学习任务，也是快速原型化科学研究中定制应用程序的平台。
引用
@article{arxiv.2411.15221,
title = {Reflections from the 2024 Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry},
author = {Yoel Zimmermann and Adib Bazgir and Zartashia Afzal and Fariha Agbere and Qianxiang Ai and Nawaf Alampara and Alexander Al-Feghali and Mehrad Ansari and Dmytro Antypov and Amro Aswad and Jiaru Bai and Viktoriia Baibakova and Devi Dutta Biswajeet and Erik Bitzek and Joshua D. Bocarsly and Anna Borisova and Andres M Bran and L. Catherine Brinson and Marcel Moran Calderon and Alessandro Canalicchio and Victor Chen and Yuan Chiang and Defne Circi and Benjamin Charmes and Vikrant Chaudhary and Zizhang Chen and Min-Hsueh Chiu and Judith Clymo and Kedar Dabhadkar and Nathan Daelman and Archit Datar and Wibe A. de Jong and Matthew L. Evans and Maryam Ghazizade Fard and Giuseppe Fisicaro and Abhijeet Sadashiv Gangan and Janine George and Jose D. Cojal Gonzalez and Michael Götte and Ankur K. Gupta and Hassan Harb and Pengyu Hong and Abdelrahman Ibrahim and Ahmed Ilyas and Alishba Imran and Kevin Ishimwe and Ramsey Issa and Kevin Maik Jablonka and Colin Jones and Tyler R. Josephson and Greg Juhasz and Sarthak Kapoor and Rongda Kang and Ghazal Khalighinejad and Sartaaj Khan and Sascha Klawohn and Suneel Kuman and Alvin Noe Ladines and Sarom Leang and Magdalena Lederbauer and Sheng-Lun and Liao and Hao Liu and Xuefeng Liu and Stanley Lo and Sandeep Madireddy and Piyush Ranjan Maharana and Shagun Maheshwari and Soroush Mahjoubi and José A. Márquez and Rob Mills and Trupti Mohanty and Bernadette Mohr and Seyed Mohamad Moosavi and Alexander Moßhammer and Amirhossein D. Naghdi and Aakash Naik and Oleksandr Narykov and Hampus Näsström and Xuan Vu Nguyen and Xinyi Ni and Dana O'Connor and Teslim Olayiwola and Federico Ottomano and Aleyna Beste Ozhan and Sebastian Pagel and Chiku Parida and Jaehee Park and Vraj Patel and Elena Patyukova and Martin Hoffmann Petersen and Luis Pinto and José M. Pizarro and Dieter Plessers and Tapashree Pradhan and Utkarsh Pratiush and Charishma Puli and Andrew Qin and Mahyar Rajabi and Francesco Ricci and Elliot Risch and Martiño Ríos-García and Aritra Roy and Tehseen Rug and Hasan M Sayeed and Markus Scheidgen and Mara Schilling-Wilhelmi and Marcel Schloz and Fabian Schöppach and Julia Schumann and Philippe Schwaller and Marcus Schwarting and Samiha Sharlin and Kevin Shen and Jiale Shi and Pradip Si and Jennifer D'Souza and Taylor Sparks and Suraj Sudhakar and Leopold Talirz and Dandan Tang and Olga Taran and Carla Terboven and Mark Tropin and Anastasiia Tsymbal and Katharina Ueltzen and Pablo Andres Unzueta and Archit Vasan and Tirtha Vinchurkar and Trung Vo and Gabriel Vogel and Christoph Völker and Jan Weinreich and Faradawn Yang and Mohd Zaki and Chi Zhang and Sylvester Zhang and Weijie Zhang and Ruijie Zhu and Shang Zhu and Jan Janssen and Calvin Li and Ian Foster and Ben Blaiszik},
journal= {arXiv preprint arXiv:2411.15221},
year = {2025}
}
备注
Updating author information, the submission remains largely unchanged. 98 pages total