# Paper Shows Prompt Formatting Alone Can Flip Language-Model Leaderboards

A token-controlled study introduces a Format Sensitivity Index measuring how much identical questions in different wrappers move model scores.

- Published: 2026-07-14T05:33:22.579Z
- Canonical: https://polylog.news/ai/2026-07-14/paper-shows-prompt-formatting-alone-can-flip-language-model
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv](https://arxiv.org/abs/2607.09665)

A paper posted today argues that prompt wrappers that differ only in formatting can change model scores enough to reverse leaderboard conclusions. The authors study this variance under a token-controlled protocol, holding the number of toke…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-07-14/paper-shows-prompt-formatting-alone-can-flip-language-model (subscription information: https://polylog.news/pricing).