Do Language Models Dream of Binding Molecules? Benchmarking LLMs under Spatial Constraints
The paper tests whether general-purpose LLMs can generate molecules while obeying 3D binding constraints.
The authors introduce 3D-Fit, a token-efficient benchmark for pocket-conditioned molecule generation under spatial constraints. The setup includes anchor fragments, pharmacophore points, and required pocket-ligand interactions. They find that LLMs remain behind specialized diffusion models, but can handle multiple spatial constraints at once. Source: ArXiv · AI/CL/LG's note
The authors introduce 3D-Fit, a token-efficient benchmark for pocket-conditioned molecule generation under spatial constraints. The setup includes anchor fragments, pharmacophore points, and required pocket-ligand interactions. They find that LLMs remain behind specialized diffusion models, but can handle multiple spatial constraints at once. Source: ArXiv · AI/CL/LG's note
score 4