最近axis robotics其实有一些非常amazing的进展。还没形成paper,但是先简单分享一下激动的心情。
@0xsexybanana42K followers
19,739 views
33 Comments
Comments from the original post on X (Twitter)

@axisroboticsPh
Great work from all contributors mam. We are happy we are a part of it!
2 likes

@web3rayxy
太酷了 这才是物理AI该有的样子
1 like

@onderfhm
Oh look like my tactic strategy 🤩😅
1 like

@0xmosk
子时说的是真的吗
1 like

@max_bevza
shift from “train one giant model” to “build experts, then distill them” is really interesting
1 like

@whb0821
极其清晰且务实的具身智能路径!放弃单靠海量通用数据强推的迷思,先用社区轨迹作先验,靠 LLM 自动化生成场景,再用 RL 把专项 Expert 训练到 100% 成功率。当成百上千个垂直 Expert 跑通仿真到真实世界的闭环,最后通过统一蒸馏反哺 Foundation Model,这才是工程落地最性感的解法。
1 like

@Ethan6xa7
数据飞轮滚起来后面就越来越快

@BanhanETH
Community data is the real moat if it’s diverse enough

@vanteobn
so easy

@wilfried25_
semakin di hina semakin menyala 🤟

@VasuSha01579214
opening drawer is not easy as it look...

@IsraelJIL
when do you see this franka arm? the white is catchy abit

@yudeewxy
数采可以分布式来做,数采的硬件免费给到用户,用户贡献数据,获得奖励。

@0_emes
Its vesting or no for community contributor? Airdrop no vesting right? @0xsexybanana

@ShaikAn42246985
说了一堆ai花絮话术,都避免不了,你的项目被你这个无知且愚昧的女人搞砸了!

@suqiu69775097
空气项目又来割韭菜了

@lui5Z
看着就带劲 这种实打实的进步真让人高兴

@RosePhamm2002
Bạn cảm thấy như thế nào khi nghe về những tiến bộ vô cùng tuyệt vời của Axis Robotics? 🌟

@Shayueth
专家网络跑起来记得发个演示看看

@McCain889
从社区数据筛方向、RL 批量出专项专家、再统一蒸馏,路径挺清晰的,后面专家之间能互相迁移就更强了











