Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Shishan AI Lab releases open-weight Lion-ER-35B-A3B embodied VLM

AI summary

Shishan AI Laboratory released Lion-ER-35B-A3B, an embodied reasoning vision-language model with about 35B total parameters and roughly 3B activated per token. The model team reports improvements across ten public embodied benchmarks using avg@4, with an equal-weight aggregate score of 65.29. Its BF16 Safetensors weights are available on Hugging Face and compatible with Transformers, vLLM and SGLang.

In-depth insight

The release provides developers with an open-weight embodied reasoning VLM and a capability-driven training methodology, while its benchmark results remain based on the model team's own evaluations.

允中 发自 凹非寺 量子位 | 公众号QbitAI

日常生活里人类的直觉判断,放到机器人身上,往往会变成棘手难题。

视频画面中物体位置看上去更高,并不等同于它在真实三维空间里处于更高位置;更换拍摄角度,家具外观发生改变,但它仍是同一个物体;画面内物体发生位移,有可能是物体本身发生移动,也可能仅仅是相机镜头产生转动……

这些细微的空间、状态变化,人类一眼就能分辨,但对机器人的“具身大脑”来说,却是难以攻克的难题。

视角变化混淆物体身份、误判相机运动、错估任务进度,都会直接造成机器人执行失败。

想要机器人可靠完成现实任务,模型不能仅仅做到“看见物体”,更要读懂三维空间关系、捕捉物体状态变化、研判任务所处阶段。

#Lion-ER-35B-A3B#Shishan AI Laboratory#Hugging Face#Transformers#vLLM#SGLang#Qwen3.6-35B-A3B#Qwen3.6-27B

Original · 12 min readRead the original
Group chat

No comments yet. Start the conversation.