NEW OPEN SOURCE AI LEADER REFLECTION 70B'S PERFORMANCE QUESTIONED, ACCUSED OF ‘FRAUD'

Summary from the TLDR newsletter

Reflection 70B, a variant of Meta's Llama 3.1 open source large language model trained and released by New York startup HyperWrite, has been aggressively questioned by third-party evaluators as it fails to reproduce some of its claimed performance measures.