OpenAI launches LifeVal benchmark for everyday assistant tasks
OpenAI introduced LifeVal, a benchmark that evaluates whether AI can complete practical personal-assistant workflows such as childcare, health, education, household administration, finances, shopping and government services. OpenAI says it shaped the benchmark using time-use studies and a survey of 3,995 respondents, and used it for a year to track model improvements. The company reported scores including 93.1% for kindergarten selection, 98.0% for organizing flood volunteers and 90.4% for preparing a diabetic relative’s meal plan, but LifeVal remains primarily an internal evaluation without its full task set or scoring code publicly available.
LifeVal shifts evaluation toward executable, deliverable everyday workflows, but its example scores cannot yet be independently reproduced because the complete benchmark and scoring code remain internal.
For AGI to benefit all of humanity, it should help people with the everyday tasks that matter most to them. So we built LifeVal, a new benchmark shaped by feedback from thousands of ChatGPT users on where they most want help from a personal assistant—from finding childcare to accessing government services to managing personal finances.
Over the past year, we’ve used LifeVal to measure and improve how well our models handle these everyday tasks.
https://openai.com/index/lifeval/