Datasets:

ermiaazarkhalili
/

orca-mini-short-100tok

license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - xai
  - attribution
  - decoder
  - short-context
size_categories:
  - 1K<n<10K

Orca-Mini - Short Context (<100 tokens)

This dataset is a filtered version of ermiaazarkhalili/orca-mini-v1-high-prob-qwen-0.5b-10k containing only samples with fewer than 100 tokens.

Dataset Description

This dataset is designed for efficient XAI (Explainable AI) attribution evaluation on decoder models. Short context samples allow for faster evaluation while maintaining meaningful attribution analysis.

Statistics

Total samples: 5,000
Token count range: 5 - 99
Mean token count: 23.1 (std: 11.8)
Median token count: 21.0

Columns

Column	Description
`sentence_prefix`	Text prefix before predicted token
`predicted_token`	Model's top-1 prediction
`actual_token`	Ground truth token
`probability`	Prediction confidence (0-1)

Source Dataset

This dataset was created by filtering ermiaazarkhalili/orca-mini-v1-high-prob-qwen-0.5b-10k with the following criteria:

Token length: 5 <= tokens < 100
Tokenizer: Qwen/Qwen2.5-0.5B

Usage

from datasets import load_dataset

dataset = load_dataset("ermiaazarkhalili/orca-mini-short-100tok", split="train")
print(f"Loaded {len(dataset)} samples")

Citation

If you use this dataset, please cite the ALTI-XAI framework.

License

MIT License