Do Language Models Dream of Binding Molecules? Benchmarking LLMs under Spatial Constraints
The paper tests whether general-purpose LLMs can generate molecules while satisfying 3D binding constraints.
The authors introduce 3D-Fit, a benchmark for pocket-conditioned ligand generation with constraints such as anchor fragments, pharmacophore points, and required pocket-ligand interactions. They compare LLM-based methods with specialized diffusion-model baselines. The abstract says LLMs still trail state-of-the-art approaches, but can handle multiple spatial constraints at once. Source: HF Daily Papers' note
The authors introduce 3D-Fit, a benchmark for pocket-conditioned ligand generation with constraints such as anchor fragments, pharmacophore points, and required pocket-ligand interactions. They compare LLM-based methods with specialized diffusion-model baselines. The abstract says LLMs still trail state-of-the-art approaches, but can handle multiple spatial constraints at once. Source: HF Daily Papers' note
score 4