AI Benchmarking Platform LM Arena Get Talk Say Dem Dey Show Favor For Big Tech Labs, Make People Worry About How Clear Everything Be

Chatbot Arena, wey UC Berkeley researchers start for 2023 and now Arena Intelligence Inc. don turn am to business, don get wahala small small since new research wey involve Cohere, Stanford, MIT, and Ai2 come out. Di study talk say LM Arena, di platform wey dey behind Chatbot Arena, dey give extra favour to big big tech labs like Meta, OpenAI, Google, and Amazon. Di accuse dem make na say dem dey allow selected private testing, only show di highest scores of models, and give di main labs AI models more head-to-head fights. Di researchers call dis style 'gamification', and dem talk say e dey spoil AI model rankings and make tin no clear. LM Arena say di research no correct and dem serious about fair benchmarking. For response, di study suggest make tin clear pass by limiting private pre-release tests, make all model scores public, and change di way dem select models to make sure everybody get fair chance. Di palava wey dey go on show say trust, clear tin, and no partiality really important for AI benchmarking, especially as LM Arena dey go into business model and di sector dey important well well for tech and investment people.