Unverified50% confidenceFactExact time
Early large language models were prone to forgetting prior context or hallucinating in long-context scenarios.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当叙事篇幅超出上下文窗口长度时，早期建立的角色设定与情节伏笔会逐渐淡出模型的有效记忆范围，导致生成内容前后矛盾76% similarUnverified模型规模越大，幻觉的表述往往越流畅、越具迷惑性，识别难度反而上升73% similarVerifiedThe attention mechanism of the Transformer architecture significantly reduces its focus on early content when processing very long sequences, causing the model to forget earlier settings73% similarUnverifiedLarge language models are prone to hallucination drift when handling detail consistency problems that require precise long-range context tracking.72% similarUnverifiedIn long contexts, models are more easily influenced by the most recent few turns of conversation rather than the system prompt set at the beginning, a phenomenon known as instruction-following decay.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897375API
curl https://kongchang.com/api/v1/knowledge/claims/897375MCP
get_claim(id=897375)