Sentient ROMA Builds Hierarchical Multi-Agent Systems With Recursive Task Trees

Sign in✓ Signed in
Sentient’s ROMA (Recursive Open Meta-Agents) is an open-source meta-agent framework that builds hierarchical multi-agent systems by recursively splitting a complex goal into subtasks and aggregating the answers back up a task tree. The project, documented on Sentient’s blog and in a February 2026 paper from Sentient with collaborators at Virginia Tech, UC Berkeley, UC San Diego, and the University of Maryland, targets long-horizon work where sequential agent loops grow brittle, overflow context, and hide the step that failed.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...