database-grounded verification

New verification method cuts LLM chemical reasoning errors from 22% to 4%

Researchers have developed a new method for verifying the accuracy of large language models (LLMs) in chemical and materials reasoning. This approach uses a tiered verifier that checks claims against authoritative datab…