OpenAI introduces LifeVal to assess models on everyday tasks

發生什麼事

@OpenAI says LifeVal was already in use: the company used it over the past year to assess and improve its models’ performance on everyday tasks. Its test cases cover childcare, government services and personal finances, reflecting what thousands of ChatGPT users look for in a personal assistant. The design and reported improvements are OpenAI’s account; this site has not tested the benchmark (🗣 self-reported).