Unverified50% confidenceFactExact time
The typical structure of the agent loop is observe→think→act→observe, and in each round the harness resends the full conversation history to the model.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn each iteration of the loop, an agent needs to repackage the system prompt, conversation history, tool descriptions, and all outputs from previous steps before feeding them into the model, causing token consumption to grow linearly or even super-linearly with the number of steps.79% similarUnverifiedAn agent is essentially a loop in code around a model: construct a prompt, hand it to the model for a decision, have code execute the tools and put the results back into the prompt, and repeat until the model produces a final answer78% similarUnverified思维链将模型内部工作记忆外部化,把中间推导结果存入上下文,借助注意力机制随时读取和复用76% similarVerifiedThe core of a ReAct agent is the "reasoning-action-observation" loop, in which the LLM analyzes the problem, decides to call a tool, observes the returned result, and continues reasoning if unsatisfied until it gives a final response75% similarVerifiedThe core idea of ReAct is to interweave the language model's thinking and actions, forming an iterative loop of Think → Act → Observe, which differs from the pure reasoning chain of Chain-of-Thought.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085802API
curl https://kongchang.com/api/v1/knowledge/claims/1085802MCP
get_claim(id=1085802)