BayesCode/

smollm3-blindspots-analysis

	SmolLM3-3B Blind Spots Dataset




	
	Dataset Description

This dataset documents failure cases (“blind spots”) observed while testing the base language model :contentReference[oaicite:0]{index=0}.
The goal of this dataset is to identify scenarios where the model produces incorrect, misleading, or unreliable outputs when responding to prompts. The dataset includes three main fields:

Input – The prompt given to the model
Expected Output – The correct or logically expected… See the full description on the dataset page: https://huggingface.co/datasets/BayesCode/smollm3-blindspots-analysis.