Multimodal agents.
One research library.

Control over generation conditions, execution, feedback, and reusable experience across agentic image, video, 3D, and multimodal generation

Observation compression, memory, action efficiency, and runtime optimization for GUI agents; connects capability evaluation with token, latency, and execution costs

Evidence audit of 336 GUI-agent papers across architecture, interactive evaluation, recovery, lifecycle engineering, privacy, observability, human oversight, and deployment readiness

Agentic internal intelligence, external tool invocation, environment interaction, training resources, evaluation, and applications for multimodal large language models

Architectures, tool use, collaboration, applications, and evaluation for LLM-driven multimodal agents

Interactive agents that integrate environmental perception, multisensory inputs, external knowledge, embodied action, and human feedback

Connects computer-use and robot-use through Perceive, Anticipate, Plan, Act, and Verify (PAPAV), with attention to physical constraints and gaps in evaluation

Phillip Isola examines general-purpose agents operating robots through sensor and actuator interfaces, including latency, task feedback, and the limits of current deployments

General Robotics describes GRID workflows for hardware integration, simulation, skill construction, and deployment repair; laboratory examples are developer reports

Evidence-audited review of 42 intelligent-transportation agent families, separating model, system, and hybrid multimodality across capability, validation, deployment readiness, and bounded orchestration

Five functional roles for language, the embodied modules that consume it, and an evidence audit spanning actions, interventions, and causal grounding claims

Contrasts end-to-end VLA weight policies with generated code skills, including closed-loop repair, skill memory and evolution, and interoperable skill ecosystems