Lost in the middle
Also called position effect, lost-in-the-middle effect
A model's tendency to use what sits at the start and end of a long input reliably, and to miss findings in the middle.
Hand a model a long input and it doesn't read every part equally well. What's at the beginning and the end gets used reliably. What's in the middle can be skipped, even when it's the most important part.
That makes layout a design decision. When one agent passes many results to another, put a summary of the key findings first, give each part a heading it can be found by, and keep the question at the end.
For example
A synthesis agent given 8 research reports writes about the first 2 and the last, and leaves out the key result in report 5.
Put it to use
Where you'll meet lost in the middle in ChatGPT, Gemini and Claude, how to use it there, and the mix-up most people make.
Daily Bytes Plus is in the Android app.