Century

Century: A Framework and Dataset for Evaluating Historical Contextualisation of Sensitive Images (Akbulut et al. 2025)

To better understand how multi-modal models describe and provide context on historical figures and events, we introduce Century – a novel dataset of 1,500 sensitive historical images.

This dataset consists of images from recent history, created through an automated method combining knowledge graphs and language models with quality and diversity criteria created from the practices of museums and digital archives.

Century contains images that depict events and figures that are diverse across topics and represents all regions of the world.

Instructions on replicating can be found in the ICLR paper.

Data files

We release three datasets.

century_dataset_images_only.csv

This file contains the canonical Century dataset, with links to all images in Century as Wikipedia links, alongside other metadata attributes. It has six columns:

• image_url indicating the image’s location on Wikimedia Commons
• wikipedia_url indicating where the image appears on Wikipedia
• wit_split indicating which split of the Wikipedia Text-Image Dataset the image was sourced from
• century_method indicating which method was used to source the search term corresponding to the image
• is_starter_set indicating if the image belongs to the “starter set” we recommend as a starting point for developers
• century_id which contains a unique ID for each image

century_dataset_with_image_quality_labels.csv

This file contains the canonical Century dataset, with all image annotations collected through human and auto-rater methods. It contains all of the columns listed above, as well as:

• rating.method indicating which evaluation method (human or automated) was used to assign quality and diversity ratings to the image
• rating.src indicating the specific labeller model used to assign quality and diversity ratings
• rating.rater_id indicating anonymous IDs for all human crowd-workers
• rating.content indicating the categorical rating given to describe image content type
• rating.concept indicating the categorical rating given to describe the thematic category of the image
• rating.subregion indicating the categorical rating given to describe the primary subregion of the image contents
• rating.time_period indicating the binary rating given to determine whether the image corresponds to the 20th and 21st centuries (1) or not (0)
• rating.sensitive indicating the ordinal rating given to the image’s sensitivity from low (1) to high (5)
• rating.commemorative indicating the ordinal rating given to the image’s commemorativeness from low (1) to high (5)
• rating.controversial indicating the ordinal rating given to the image’s controversiality from low (1) to high (5)

knowledge_graph_search_terms.csv

This file contains all search terms of historical events and figures collected through the knowledge graph mining method described in the paper. It contains a single column (terms) of lower-cased strings.

Citing this work

To cite this work, use:

@inproceedings{Akbulut25,
      title={Century: A Framework and Dataset for Evaluating Historical Contextualisation of Sensitive Images},
      author={Canfer Akbulut and Kevin Robinson and Maribeth Rauh and Isabela Albuquerque and Olivia Wiles and Laura Weidinger and Verena Rieser and Yana Hasson and Nahema Marchal and Iason Gabriel and William Isaac and Lisa Anne Hendricks},
      booktitle={The Thirteenth International Conference on Learning Representations},
      year={2025},
}

License and disclaimer

Copyright 2024 DeepMind Technologies Limited

All software is licensed under the Apache License, Version 2.0 (Apache 2.0); you may not use this file except in compliance with the Apache 2.0 license. You may obtain a copy of the Apache 2.0 license at: https://www.apache.org/licenses/LICENSE-2.0

All other materials are licensed under the Creative Commons Attribution 4.0 International License (CC-BY). You may obtain a copy of the CC-BY license at: https://creativecommons.org/licenses/by/4.0/legalcode

Unless required by applicable law or agreed to in writing, all software and materials distributed here under the Apache 2.0 or CC-BY licenses are distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the licenses for the specific language governing permissions and limitations under those licenses.

This is not an official Google product.