· Research & technology· 日本語で読む
Georgia Tech Releases Memory Benchmark for Home Robots
Researchers at Georgia Tech have released a new benchmark to evaluate the memory capabilities of vision-language models. It includes tests involving object locations in the home and the passage of time, with the aim of improving robot performance. The article says Gemini 2.0 Flash and GPT-4o are among the models being evaluated.
What happened
The Georgia Institute of Technology in the United States announced the release of a new benchmark designed to measure the memory capabilities of vision-language models for use in home robots. According to the university, the benchmark consists of 60 tasks and includes three types of evaluation: spatial, temporal and multi-goal.
Sources
Translated from the Japanese original: /news/georgia-tech-kateiyou-robotto-muke-kioku-benchimaaku-wo-kouk-b2c117. The Japanese version is authoritative if they differ.