AITOPAI热报
ModelsMulti-source

清华 × 伯克利开源 uni-robot-agent，用 Qwen3.8-Flash-Next 让大模型指挥机器人

AIGCLINK@aigclink
Post· 11 sources
View the post
Why we picked it

清华和伯克利开源了个机器人 Agent 壳，套上 Qwen3.8-Flash-Next 就在 RoboDojo 上赢了 GPT-6 Astra，成本还只有 1/20，搞机器人的可以看看。

uni-robot-agent 是清华与伯克利合作的开源机器人 Agent harness，不依赖仿真器真值，也不做蒸馏，直接给开源模型 Qwen3.8-Flash-Next 套壳让它指挥真机器人。在 RoboDojo 基准上拿到 31.39 分，超过 GPT-6 Astra 的 28.97，单局成本约为 GPT 的 1/20。在 19 个真机任务上，也超过了微调过的 π0.5 和 zero-shot 的 Dream-1。该项目对标宇树与 GPT-6 Astra 结合的路线。

Original · AIGCLINK

uni-robot-agent：清华 × 伯克利一起做的开源机器人 Agent harness，给大模型llm套一层"壳"，让它直接去指挥机器人干活，对标最近宇树和gpt-6 astra结合指挥机器人干活的路径。

1、没用仿真器真值，也没做蒸馏，就是用现成的开源模型 Qwen3.8-Flash-Next 往上一套harness。 2、开源模型反超闭源：RoboDojo 基准上拿了 31.39，GPT-6 Astra 是 28.97，而且每局成本大概只有gpt的 1/20。 3、真机测试：19 个真机任务上，也超过了微调过的 π0.5 和 zero-shot 的 Dream-1。

一句话：机器人上面那层物理AI"大脑"，现在不一定非得靠闭源大模型+昂贵后训练了，harness 做好了，开源模型也能打。 #物理ai #robotagent

#uni-robot-agent#Qwen3.8-Flash-Next#GPT-6 Astra#物理AI#开源模型

Keep reading

Recent stories with the same tag or section
PapersBuildersMulti-sourceSelected10:33
Tsinghua and Berkeley open-source uni-robot-agent framework

The key insight is that framework design, not model training, drives most of the zero-shot gain: reliable tool feedback and consistent interfaces lifted the same model over 16x on RoboDojo. The five-layer stack plus six reliability rules (snapshot binding, queue flush, pose-error reporting) is a practical blueprint for anyone building real-robot or simulation agents. Simulation and real-machine interaction logs can be combined for later training, and the code is open-source on Hugging Face.

#uni-robot-agent#Qwen#DeepSeek#open-source models
Event
shao__meng@shao__meng11 sourcesSource
ModelsBuildersMulti-sourceSelected01:14
VIALS-AA leaderboard ranks models on life science visual reasoning

The 29× cost gap between frontier models with a 3-point score spread changes the economics of choosing a model for scientific agent workloads. Teams building lab-automation or biotech analysis tools can use the per-task cost and turn counts to pick models that fit their budgets. The 19-point deficit of open weights models like Kimi K3.1 Pro also marks a concrete ceiling for on-premise deployments.

#VIALS#Gemini 4 Carbon#GPT-6#Claude Opus 5.5
Event
官方账号Artificial Analysis@ArtificialAnlys11 sourcesSource