Peking University, in collaboration with Shanghai Upperway Quintel Research Institute, released the benchmark dataset for home service robot user habits.

Peking University, in collaboration with Shangwei New Materials Shangwei Qiyuan Research Institute, recently released the UcON dataset. It is reportedly the world's first benchmark for individualized navigation evaluation based on user's personalized lifestyle habits. The benchmark covers 489 types of objects and 22,600 habit data, constructed with differentiated scenarios through LLM generation + physical verification. Experiments show that its habit retrieval mechanism improves navigation efficiency by 35%.