We track how language models handle facts with our March 2026 update. We test...

https://www.mediafire.com/file/cjr04fsf8o7uktx/pdf-58192-83696.pdf/file

We track how language models handle facts with our March 2026 update. We test top models against the FACTS benchmark to measure accuracy and reliability. Our research shows that leading systems now hold a 0.7% hallucination rate on verified corporate data