Identifying Latent Actions and Dynamics from Offline Data via Demonstrator Diversity

ArXi:2603.17577v1 Announce Type: new Can latent actions and environment dynamics be recovered from offline trajectories when actions are never observed? We study this question in a setting where trajectories are action-free but tagged with nstrator identity. We assume that each nstrator follows a distinct policy, while the environment dynamics are shared across nstrators and identity affects the next observation only through the chosen action.