Lambda reports EGInterpolator reduces molecular dynamics gaps on DRUGS

Lambda said AI for molecular dynamics is constrained by expensive training trajectories, and presented EGInterpolator as ICLR 2026 work with Stanford. The method first learns molecular structure from abundant conformer data, then uses scarce MD data to learn motion.

On the DRUGS benchmark, Lambda said EGInterpolator reduced the gap to reference simulations by 73% for bond angles, 78% for bond lengths, and 24% for torsional motion.

A structure-first ablation was also reported: removing pretraining increased mean JSD from 0.173 to 0.332 for bond angles and from 0.142 to 0.386 for bond lengths.