From Benchmark to Bench: Can Agents Survive Real-World Drug Discovery?
Pierre Llompart, Levent Güner, Helen Lai, Alessandro Tibo, Yijie Xu
The paper evaluates an agentic system for drug discovery, finding that its performance depends on predictive models rather than generation methods, and that outputs are indistinguishable from expert work.
bioRxiv · Oct 6, 2026 · PreprintBioinformatics