Preference Optimization for Molecular Language Models

Ryan Park, Ryan Theisen, Navriti Sahni, Marcel Patek, Anna Cichońska +1 more
10/18/2023
stat.MLcs.AIcs.LG

No official code found. Best implementation: Harmonic-Discovery/pref-opt-for-mols (7 stars, last commit 2024-07-23).

Abstract

Molecular language modeling is an effective approach to generating novel chemical structures. However, these models do not \emph{a priori} encode certain preferences a chemist may desire. We investigate the use of fine-tuning using Direct Preference Optimization to better align generated molecules with chemist preferences. Our findings suggest that this approach is simple, efficient, and highly effective.

View on arXivView PDF

Code Implementations(1)

Harmonic-Discovery/pref-opt-for-molsCommunity65%

Supporting code for "Preference Optimization for Molecular Language Models"

72Python, Jupyter NotebookSep 26, 20232 years ago

Similar Papers

Small Molecule Optimization with Large Language Models

87% match1 repo372024

Domain-Agnostic Molecular Generation with Chemical Feedback

85% match1 repo1952026

Domain-Agnostic Molecular Generation with Chemical Feedback

85% match1 repo1952023

Text-guided multi-property molecular optimization with a diffusion language model

85% match1 repo42024

Large Language Models for Controllable Multi-property Multi-objective Molecule Optimization

84% match1 repo52025

Efficient Evolutionary Search Over Chemical Space with Large Language Models

84% match1 repo572024

Cite this paper

@article{park2023preference,
  title  = {Preference Optimization for Molecular Language Models},
  author = {Ryan Park and Ryan Theisen and Navriti Sahni and Marcel Patek and Anna Cichońska and Rayees Rahman},
  year   = {2023},
  eprint = {2310.12304},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2310.12304v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2310.12304.svg)](https://codeofpaper.com/paper/2310.12304)

Discussion