Unverified50% confidenceFactExact time
Mobile agents typically use multimodal large models (MLLMs) as the decision-making core, with a typical architecture comprising three stages—perception, reasoning and planning, and action execution—often chained together as a 'reasoning-to-action' pipeline.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMultimodal large language models (MLLMs) are large-scale neural network models capable of simultaneously processing and understanding multiple modalities such as text, images, audio, and video, typically based on the Transformer architecture, and acquiring cross-modal alignment capabilities through pretraining on large-scale multimodal data.74% similarUnverifiedIn an agent, the large model is responsible for reasoning and planning decisions, while tools are responsible for executing tasks, with tool invocation uniformly controlled and decided by the large model.73% similarVerifiedReAct requires the model to explicitly output a "Thought—Action—Observation" triplet at each step, forming a traceable decision trajectory.70% similarUnverified模型直觉包括判断任务是否需要 RAG、Agent 多步推理或微调等技术路径的选择能力70% similarUnverifiedThe core idea of 'Jeva-class' models is to compress complex multi-step reasoning into rapidly executable decision paths, used in agent scenarios such as tool invocation, task assignment, and routing.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069461API
curl https://kongchang.com/api/v1/knowledge/claims/1069461MCP
get_claim(id=1069461)