arXiv:2606.03057v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for molecular tasks, but it remains unclear which molecular representation to use. We present a systematic benchmark evaluating LLM molecular competence across nine representations and eight chemical tasks. We benchmark 16 LLMs across five model families, including reasoning and non-reasoning variants, chemistry-specialized LLMs, and closed frontier models. Performance is strongly representation-dependent and no single representation wins across tasks, though CML is the best, followed by MolJSON,
The proliferation of Large Language Models (LLMs) into specialized domains like molecular science necessitates a systematic understanding of their optimal representations and capabilities.
This research provides crucial insights into how LLMs can effectively process and reason about molecular data, directly impacting drug discovery, material science, and synthetic biology applications.
The empirical study clarifies the performance dependencies of LLMs on molecular representations, guiding future development and application in chemistry-related tasks.
- · AI researchers in chemistry
- · Pharmaceutical companies
- · Materials science
- · Synthetic biology
- · Traditional molecular modeling methods without AI integration
- · LLM developers ignoring representation optimization
More efficient and accurate molecular design and prediction using LLMs.
Accelerated discovery of new drugs, materials, and biological pathways.
Enhanced automation in R&D leading to faster product cycles and novel industries.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG