OpenAIAnalysisAI ModelsNovember 22, 2026

OpenAI argues perplexity should be a shared AI progress metric

Read original source →openai.com

OpenAI published internal pretraining perplexity measurements, arguing cross-entropy loss should become an industry-wide yardstick for comparing model developers and informing shared safety standards. The post says publishing the numbers gives competitors a clearer picture of where OpenAI stands, a tradeoff it says is outweighed by the need for shared safety measures.

2 sources

More stories today

Open the live feed