ProductsMulti-sourceSelected

Hedra 推出 Model Factory,把 RSI 闭环开放给机器人等领域团队

Why we picked it

Hedra 把自家 AI 科学家跑 post-training 的流程做成了服务,你带模拟器和 RL 环境进去就行。他们在 H3 视频模型上先跑通,Physics-IQ 从 30.3 提到 34.9。做机器人、制造或国防模型、又没有内部训练团队的工程团队可以看看。

Hedra 推出 Model Factory,一个自主推理与训练平台,面向机器人、制造、生物和国防领域团队。团队带上自己的模拟器或 RL 环境,Hedra 的 AI 科学家会执行 post-training,Hedra 研究人员负责引导整个过程。

Body

平台先在 Hedra 自家的 H3 视频模型上验证:改进后版本在 Physics-IQ 基准上拿到 34.9 分,相比 H3 Max 的 30.3 分提升了 4.6 分。这套递归自我改进(RSI)闭环此前只存在于少数前沿实验室内部,Hedra 把它拆成服务对外开放。

Hedra 团队成员来自 Stanford、NVIDIA 和 Meta Superintelligence Labs,走的是研究人员与客户并肩工作的模式,而不是纯自助 API。对没有内部 post-training 团队的垂直行业公司来说,这提供了一条把领域模拟器转化为模型改进的路径。

Original · elvis

RSI loops are moving out of the frontier labs. @hedra_labs just launched Model Factory for teams in robotics, manufacturing, biology, and defense. Bring a simulator or RL environment, and their AI scientists run post-training while Hedra researchers steer. They first tested it on their own H3 video model. 34.9 on Physics-IQ vs. 30.3 for H3 Max. Hedra @hedra_labs Introducing Hedra Model Factory: an autonomous inference and training platform for building models for the real world. The recursive self-improvement loop behind frontier research has been locked inside a few labs. We're opening it to teams in robotics, manufacturing, biology and defense. Our researchers come from Stanford, NVIDIA and Meta Superintelligence Labs. Now we work alongside your team to post-train and optimize your models.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Hedra, RSI, post-training, Physics-IQ.

Source attribution: curated from elvis (@omarsar0). Cross-source corroboration: 11 independent sources have covered this development (including Together AI, Meta Engineering Blog, NVIDIA AI), confirming industry-wide signal strength.