Papers & Research

Meituan LongCat Open-Sources VitaBench 2.0: A New Benchmark for Long-Term Dynamic Agents

AI summary

The Meituan LongCat team open-sourced VitaBench 2.0, the first agent evaluation benchmark for long-term dynamic user modeling in real-life scenarios, including 56 users with real characteristics, 819 complex tasks, over 2,000 dynamic preferences and 66 executable tools, with an average of 2,093 interaction events per user spanning 1,580 days.

Key Takeaways

  • The Meituan LongCat team open-sourced VitaBench 2.
  • the first agent evaluation benchmark for long-term dynamic user modeling in real-life…
  • 819 complex tasks,000 dynamic preferences and 66 executable tools,with an average of 2,093…

Quick Facts

Published
Updated
Source
公众号:龙猫LongCat(美团)
Category
Papers & Research
AI score
69
Read the original Read the full article at the original source
Why it matters

它将评测重心从单次任务转移到对用户偏好的长期追踪与主动沟通,实验显示主动提问和偏好利用仍是普遍短板,记忆策略并非简单叠加,这会改变智能体个性化研究的评估方法。

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Meituan LongCat Open-Sources VitaBench 2.0: A New Benchmark for Long-Term Dynamic Agents. June 25, 2026. https://new.nat6.net/en/news/cmsdu2t2a011lrofzwicu7fny/

View all →