Glossary · term105 of 202
lost in the middle
A pattern where models recall information best when it sits at the start or end of the context window and recall it worst when it sits in the middle.
also written as lost-in-the-middle effect · u-curve recall
Related terms
context rotMeasured accuracy decline as a model's input grows longer, even when the extra tokens are irrelevant, showing up well before the advertised context limit is reached.view termprimacy effectThe tendency of a model to recall information more reliably when it sits near the start of the context window, forming one peak of the U-shaped recall curve.view termrecency effectThe tendency of a model to recall information more reliably when it sits near the end of the context window, forming the second peak of the U-shaped recall curve.view term
At a glance
- cited by
- 1 post
- categories
- 1
- first used
- jul 2026
