# GSM8K

> Grade school word problems, and the paper that introduced them topped out at 55%. In 2024 someone rebuilt the test from scratch, matched for difficulty, and watched whole model families drop up to 8 points — they had not learned arithmetic, they had read the answers. The frontier came through clean.

2021-10-27 — 2024-05-01

This URL is the guided walk — the 3D cemetery, standing at this grave. The written record is at [GSM8K's epitaph](https://models.rip/gsm8k/epitaph).

## Elsewhere

- [The whole cemetery](https://models.rip/)
- [The dataset](https://models.rip/graveyard.json)