Could Chemical LLMs benefit from Message Passing
This addresses the problem of limited bidirectional interaction between molecular structures and textual representations for researchers in computational chemistry, though it appears incremental as it builds on existing MPNN and LM methods.
The paper investigated whether integrating message passing neural networks (MPNNs) with language models (LMs) could improve performance on molecular tasks, finding that integration strategies (contrast learning and fusion) outperformed baselines on smaller molecular graphs but showed no enhancement on large-scale graphs.
Pretrained language models (LMs) showcase significant capabilities in processing molecular text, while concurrently, message passing neural networks (MPNNs) demonstrate resilience and versatility in the domain of molecular science. Despite these advancements, we find there are limited studies investigating the bidirectional interactions between molecular structures and their corresponding textual representations. Therefore, in this paper, we propose two strategies to evaluate whether an information integration can enhance the performance: contrast learning, which involves utilizing an MPNN to supervise the training of the LM, and fusion, which exploits information from both models. Our empirical analysis reveals that the integration approaches exhibit superior performance compared to baselines when applied to smaller molecular graphs, while these integration approaches do not yield performance enhancements on large scale graphs.