Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
To resolve grounded control in interactive 3D environments, the authors introduce A.D.A.M.O., an embodied agent architecture that couples a tool-calling vision-language model with a synchronized visual-symbolic world model. This dual representation addresses the persistent gap between high-level natural language instruction and fine-grained spatial execution for virtual humans. Evaluated across a diagnostic taxonomy of procedural and linguistic complexity, results reveal that semantic labeling successfully eliminates perceptual ambiguity, shifting the primary failure mode downstream to physical execution while cognitive reasoning errors remain negligible.
Embodied agent bottlenecks in 3D environments stem from downstream physical execution rather than high-level reasoning once visual-symbolic grounding eliminates perceptual ambiguity.
Creating believable vh requires the coherent integration of perception, reasoning, and action mediated by language. A central challenge is to combine these components into a control loop grounded in interactive 3D environments. To this end, we present A.D.A.M.O. (Agent for language-Driven Actions with Multimodal Observations), a visual-symbolic framework for language-driven vh that leverages a pretrained vlm with tool calling to unify perception, reasoning, and action within a single control loop. A.D.A.M.O. maintains a dual visual-symbolic world model that combines egocentric visual input and synchronized symbolic state to support grounded task-oriented behavior from natural language prompts. To support diagnostic evaluation, we introduce a controlled task suite organized by a cd taxonomy that breaks down spatial tasks into procedural and linguistic complexity. Experiments in controlled scenes show that semantic labeling strongly influences task completion and failure modes, reducing perceptual ambiguity while shifting failures toward downstream execution, whereas reasoning errors remain comparatively rare.