---
license: apache-2.0
---

<b>Odd1sOut</b>: given a set of words, which one doesn't belong?

Odd1sOut is an excellent dataset for evaluating Small Language Models (SLMs) on that that larger Language Models (LLMs) usually excel at. Provided a dataset of four to six unique but sensible words, the goal is to identify which word would not fit, given the context.

Train (split): 1435 samples
Test (split): 131 samples

Example:

slip, glide, stagger, slide
ANSWER: stagger