Unverified50% confidenceFactExact time
LLM-based agents exhibit a long-horizon deceleration phenomenon in long-horizon MLE tasks: as the execution time of a single session increases, the agent's reasoning intensity gradually diminishes and task progress slows.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified研究表明LLM在处理过长、过复杂的指令集时整体遵循率会系统性下降，即指令遗忘现象77% similarUnverifiedAn overly long context dilutes the model's attention to the current task, and trimming the context often leads to more focused and accurate AI output.75% similarUnverifiedThe longer the context length, the greater the KV Cache usage, and inference speed typically decreases.75% similarUnverified对于智能体类应用，长时间稳定运行比运行速度更重要，因中断会导致缓存状态丢失和任务失败74% similarUnverified单一智能体面临上下文窗口限制和注意力稀释两大瓶颈，任务复杂度超过单次推理能力时输出质量会显著下降74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001121API
curl https://kongchang.com/api/v1/knowledge/claims/1001121MCP
get_claim(id=1001121)