Unverified50% confidenceFactExact time
When the context approaches its limit, the model may forget critical early information or produce lower-quality output.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedNative large models are constrained by context window size and tend to forget early information when handling ultra-long tasks.79% similarUnverified近似遗忘通过梯度上升等方法主动削弱特定数据对模型参数的影响，但效果难以量化验证78% similarVerifiedThe closer an event is to the cutoff date, the smaller its share in the training corpus, and the less clearly the model has mastered it.77% similarUnverifiedResearch has found that large models perform best when key information is located at the beginning or end of the context, and performance drops significantly when it is in the middle, with accuracy declining by more than 20 percentage points on certain tasks.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/996105API
curl https://kongchang.com/api/v1/knowledge/claims/996105MCP
get_claim(id=996105)