Meta's Muse Spark 1.3 System Card Uses Four Tübingen Benchmarks

Meta's system card for Muse Spark 1.3 cites four evaluation benchmarks from the Tübingen team to report the model's safety performance, drawing attention to the use of third-party benchmarks in frontier model safety assessments.

2026-10-27 ~ 2026-10-27 · 2 related posts