Dec 7, 08:32 PM

Cohere AI Launches Global-MMLU Benchmark Addressing Biases in Multilingual AI Across 42 Languages

Cohere AI has launched INCLUDE, a comprehensive multilingual language understanding benchmark aimed at addressing cultural and linguistic biases in AI evaluations. The initiative, known as Global-MMLU, spans 42 languages and seeks to provide equitable evaluation across diverse contexts. This benchmark is designed to correct existing imbalances in multilingual AI assessments by introducing a dataset that reflects global cultural sensitivities. Researchers from ByteDance have also released FullStack Bench and SandboxFusion, tools for evaluating large language models (LLMs) in real-world programming scenarios, further contributing to advancements in AI evaluation methodologies.

#AI #IN #ByteDance #FullStack Bench #SandboxFusion

Written with ChatGPT (GPT-4o mini).