Unverified50% confidenceFactExact time
A lack of information-gain awareness leads to low sample efficiency in LLMs; with the same number of interactions, human inquirers can pin down the underlying rules faster.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
Unverified模型在大海捞针测试中对中间位置信息的检索能力明显弱于首尾位置，这被称为注意力沉降现象75% similarUnverified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75% similarUnverifiedReasoning quality degrades as information accumulates in the context window, causing skill package execution to become brittle in long-horizon tasks.74% similarUnverifiedDSA稀疏注意力不仅更快，在某些任务上甚至更准，因为跳过无关token让模型更专注于有效信息74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1047443API
curl https://kongchang.com/api/v1/knowledge/claims/1047443MCP
get_claim(id=1047443)