Researchers debate whether a paper really cites IMLE that early, then shift to search compute costs

YouJiacheng · x · 2026-08-04

A thread pushes back on the claim that the paper “cites IMLE within the first two paragraphs of the approach section.” The author argues that this is misleading because there are 5.5 pages of intro/background before the method section, and that those pages are a normal place for prior work.

They also note that the first mention of IMLE is just a bracketed citation [33–35] without naming it, and that the first explicit mention of IMLE [34] appears only on page 31. In the reply, the discussion shifts to a more substantive point: search adds extra compute, so its compute-scaling properties deserve study.

Related event: IMLE Perspectives, Search Generalization, and Citation Controversy(5 posts)→

Original post →

More from Research

Research channel →