We show that our tuned LM can outperform frontier models (incl. retrieval) on the task, and retains high steerability via prompting. arxiv.org/abs/2512.13668 - work lead by the legendary @guoqingliu.bsky.social ! @maziarz.bsky.social @msftresearch.bsky.social
Solving computer-aided synthesis planning is essential for enabling fully automated, robot-assisted synthesis workflows and improving the efficiency of drug discovery. A key challenge, however, is bri...
