hendel-merullo-2023-complementary
IN premise — summaries/2026/08/24/hendel-2023-icl-task-vectors-s7-related-work.md
Created 2026-08-25T02:58:04+00:00
Merullo et al. (2023) independently demonstrated that a single vector can encode a learned function; Hendel et al. state the findings are complementary.
Summary
Two separate research groups independently showed that a single vector in a neural network can capture a learned function, and they frame their results as fitting together rather than competing. This matters because independent corroboration makes the core claim much harder to dismiss as an artifact of one lab's method, and the fact that the two groups are looking at complementary angles suggests the picture is converging from multiple directions.