gpt-6IMAGE
字节 SenseNova 评测 GPT-6 Astra 具身智能
字节 SenseNova 团队用 12 个基准、8 台真实机器人和一个新基准原型,检验 GPT-6 Astra 的具身智能边界:通用能力大体达标,实用层面还不行。
DailyPapers
@HuggingPapers9 天前收录原推发布于 2026.10.20

DailyPapers @HuggingPapers原推发布于 2026.10.20
Has GPT-6-Astra achieved embodied intelligence? Largely yes. But practical embodied intelligence? Not yet. This report from ByteDance's SenseNova team probes its limits across 12 benchmarks, 8 real robots, and a next-gen benchmark prototype.
x.com/huggingpapers/status/2112339158345658699采集于 2026.10.21 10:07