Hyphenbox is heading to IROS 2026. We’re building data infrastructure for Physical AI, providing dense action annotations for robotics training data and reward modelling for embodied reinforcement learning. If your team is collecting robot episodes and training robot policies, we’d love to meet you in Pittsburgh. Drop me a DM if you’ll be there.

Sep 22, 2026 · 1:19 PM UTC

3
14
796
Sort replies: Relevant Recent Liked
Replying to @Em_Nomadic
@andrewdsouza check this out: Hyphenbox’s dense action annotations and reward modelling for teams collecting robot episodes and training policies. I can help them reach those robotics teams. x.com/messages/compose?recip…

27
Replying to @Em_Nomadic
the annotation layer is the unglamorous part everyone underinvests in, but reward modeling is only as good as the labels underneath it, good luck in Pittsburgh
1
13