Unverified50% confidenceDecision RuleExact time
If state-space coverage stagnates over time while task success rate does not rise, it often means the agent has fallen into repetitive loops, and the weight of the exploration reward or the granularity of the memory mechanism needs adjustment.
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAs task chains grow longer, the capacity limitations of the context window and attention decay affect the model's ability to accurately maintain the initial goal, which is one of the root causes of the 'goal drift' phenomenon in long-horizon tasks.79% similarUnverifiedLong-running agents do not fail at the starting point but instead undergo chronic degradation over time in the form of latency drift, memory growth, and retrieval decay.78% similarUnverifiedFor agent-type applications, long-term stable operation is more important than execution speed, because interruptions lead to loss of cached state and task failure.78% similarUnverifiedIn experiments where time and distance were held constant, participants who passed through a doorway responded more slowly and were more likely to forget what they were holding, and passing through multiple doorways further exacerbated memory impairment.75% similarUnverified在真实使用场景中反复遇到某个知识点，比集中死记硬背更能形成持久记忆，与间隔效应和必要难度理论相印证74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073886API
curl https://kongchang.com/api/v1/knowledge/claims/1073886MCP
get_claim(id=1073886)