chatgpt-fragmentation-inconsistency-46-9pct

IN premise — summaries/2026/08/24/xie-2024-chameleon-sloth-sR-references.md

Created 2026-08-25T02:59:00+00:00

When a single piece of evidence is split into fragments, ChatGPT changes its preference in approximately 46.9% of cases (53.1% consistency rate).

Summary

Splitting a single piece of evidence into smaller pieces causes ChatGPT to flip its judgment nearly half the time, meaning you get a different answer depending on how the information is chunked rather than what the information actually says. This is a serious reliability gap for any system that processes documents in fragments, because the output depends more on formatting choices than on the underlying content.