Familiar skills, new compositions: ACG-Bench and AE-VLA
Our new research blog explores arm-wise compositional generalization, with simulation and real-robot demonstrations.
News / Field notes
New systems, research milestones, conversations, and moments that move the work forward.
Our new research blog explores arm-wise compositional generalization, with simulation and real-robot demonstrations.
Two steps toward compositional multi-arm systems and spatially object-aware manipulation policies.
Think3D is featured at VALSE 2026 in Wuhan, highlighting our work on spatial reasoning.
A lecture on open, large-scale agent society simulation and the research it enables.
Interaction rewards enable language-model agents to continuously co-evolve without external supervision.
Zaibin joins the University of Oxford as a visiting Ph.D. student.
An open social simulation platform supporting interactions among more than one million language-model agents.
Recognition for a systematic psychological perspective on multi-agent attack, evaluation, and defense.
A conversation on emerging agent systems and their broader impact.