MineExplorer: First Minute-Scale Long-Horizon Task Benchmark for Minecraft Released
Original title:让AI离开温室,走向动态世界:MineExplorer揭示顶级多模态大模型被忽视的能力断层
AI summary
Meituan's LongCat team released MineExplorer, the first benchmark for minute-scale long-horizon tasks in the open world of Minecraft, containing 813 human-verified instances. Evaluating 18 top multimodal large models found that the strongest model, Claude-Opus-4.6, achieved only a 41% overall task success rate, dropping sharply from 77% on 1-hop tasks to 12% on 4-hop tasks, with nearly 60% of failures stemming from navigation failures. MineExplorer is fully open-sourced.
Key Takeaways
- Meituan's LongCat team released MineExplorer,the first benchmark for minute-scale long-horizon…
- Evaluating 18 top multimodal large models found that the strongest model,Claude-Opus-4.
- 6,achieved only a 41% overall task success rate,dropping sharply from 77% on 1-hop tasks to 12% on…
- MineExplorer is fully open-sourced.
Quick Facts
- Published
- Updated
- Source
- 公众号:龙猫LongCat(美团)
- Category
- Papers & Research
- AI score
- 68
18 个顶级多模态模型在 Minecraft 里全翻车,每多一层隐藏前置条件成功率就砍半,搞具身智能的该看看感知和行动之间还隔着什么。
Related entities
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.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: MineExplorer: First Minute-Scale Long-Horizon Task Benchmark for Minecraft Released. July 23, 2026. https://new.nat6.net/en/news/cmryewabt00mnrolg1nxcjf18/