Yumna-a0/

qwen3-0_6b-blind-spots

This dataset contains 11 prompts where the base model unsloth/Qwen3-0.6B-Base produces incorrect or undesirable outputs. Each row has:

tag: rough failure category (math, factual, reasoning, instruction, memory, safety, lexical) input: the prompt given to the model expected: the answer a knowledgeable human would give model_output: the raw model output from the model

I used Google Colab with a… See the full description on the dataset page: https://huggingface.co/datasets/Yumna-a0/qwen3-0_6b-blind-spots.