Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2407.19089 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913448609185792 |
|---|---|
| author | Moayedpour, Saeed Corrochano-Navarro, Alejandro Sahneh, Faryad Noroozizadeh, Shahriar Koetter, Alexander Vymetal, Jiri Kogler-Anele, Lorenzo Mas, Pablo Jangjou, Yasser Li, Sizhen Bailey, Michael Bianciotto, Marc Matter, Hans Grebner, Christoph Hessler, Gerhard Bar-Joseph, Ziv Jager, Sven |
| author_facet | Moayedpour, Saeed Corrochano-Navarro, Alejandro Sahneh, Faryad Noroozizadeh, Shahriar Koetter, Alexander Vymetal, Jiri Kogler-Anele, Lorenzo Mas, Pablo Jangjou, Yasser Li, Sizhen Bailey, Michael Bianciotto, Marc Matter, Hans Grebner, Christoph Hessler, Gerhard Bar-Joseph, Ziv Jager, Sven |
| contents | Large Language Models (LLMs) have demonstrated great performance in few-shot In-Context Learning (ICL) for a variety of generative and discriminative chemical design tasks. The newly expanded context windows of LLMs can further improve ICL capabilities for molecular inverse design and lead optimization. To take full advantage of these capabilities we developed a new semi-supervised learning method that overcomes the lack of experimental data available for many-shot ICL. Our approach involves iterative inclusion of LLM generated molecules with high predicted performance, along with experimental data. We further integrated our method in a multi-modal LLM which allows for the interactive modification of generated molecular structures using text instructions. As we show, the new method greatly improves upon existing ICL methods for molecular design while being accessible and easy to use for scientists. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_19089 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Many-Shot In-Context Learning for Molecular Inverse Design Moayedpour, Saeed Corrochano-Navarro, Alejandro Sahneh, Faryad Noroozizadeh, Shahriar Koetter, Alexander Vymetal, Jiri Kogler-Anele, Lorenzo Mas, Pablo Jangjou, Yasser Li, Sizhen Bailey, Michael Bianciotto, Marc Matter, Hans Grebner, Christoph Hessler, Gerhard Bar-Joseph, Ziv Jager, Sven Computation and Language Artificial Intelligence Large Language Models (LLMs) have demonstrated great performance in few-shot In-Context Learning (ICL) for a variety of generative and discriminative chemical design tasks. The newly expanded context windows of LLMs can further improve ICL capabilities for molecular inverse design and lead optimization. To take full advantage of these capabilities we developed a new semi-supervised learning method that overcomes the lack of experimental data available for many-shot ICL. Our approach involves iterative inclusion of LLM generated molecules with high predicted performance, along with experimental data. We further integrated our method in a multi-modal LLM which allows for the interactive modification of generated molecular structures using text instructions. As we show, the new method greatly improves upon existing ICL methods for molecular design while being accessible and easy to use for scientists. |
| title | Many-Shot In-Context Learning for Molecular Inverse Design |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2407.19089 |