AI & Tech··1 sources verified
LLM Agents Lift Crystal Structure Recovery From X-ray Diffraction From 27% to 69%
Summarised by Relevant News AI · Read time: 3 min
Researchers frame powder X-ray diffraction (pXRD) structure determination as a verifiable agentic search problem, where candidate structures are scored by forward-simulated profile agreement. They introduce pXRD-Bench, 200 simulated single-phase problems with controlled distribution shifts, and report that agent adaptation raises structural recovery from 27% to 69%, with further gains from repeated sampling and verifier-guided inference-time scaling. A reflexive evolution framework then trains agents on self-generated synthetic problems and improves profile agreement on 15 measured solid-state synthesis patterns without needing solved experimental data.
Why it matters: It shows a concrete pattern for applying LLM agents with automatic verifiers to scientific tasks, which matters for teams building autonomous self-driving labs or similar agentic systems.
All sources