c2023-in-context-editing-best-ripple-edits

IN premise — summaries/2026/08/24/cohen-2023-ripple-effects-s6-conclusion-and-discussion.md

Created 2026-08-25T02:57:57+00:00

A simple in-context editing method achieves the best overall results on the RIPPLE EDITS benchmark, outperforming parametric methods ROME and ICE

Summary

A straightforward editing approach that works within the model's context window actually beats the heavier, more engineered parametric methods (like ROME and ICE) when measured on the RIPPLE EDITS knowledge-editing benchmark. This matters because it suggests that for the task of correcting or updating what a model "knows," you don't need to dig into its internal weights; a simpler, less invasive strategy already gets the best results.