Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

BlueLens Logo

Dataset Card for BlueLense

Dataset Summary

Dataset Details

Model Description Checkpoint Used Dataset Split
GDINO Features from COCO 2017 training set groundingdino-swint-ogc COCO COCO_TRAIN
GDINO Features from COCO 2017 validation set groundingdino-swint-ogc COCO COCO_VAL
GDINO ~100 random samples from COCO 2017 groundingdino-swint-ogc COCO COCO_MINI
DETR ~100 random samples from COCO 2017 detr_r50_8xb2-150e_coco_20221023_153551-436d03e8 COCO COCO_MINI
DINO-DETR ~100 random samples from COCO 2017 dino-5scale_swin-l_8xb2-36e_coco COCO COCO_MINI
  • Creators: Intel Labs

  • Version: 1.0 (Updated: 2025-05-02)

  • License: Intel Research and Development License

  • Number of Training Samples: >100M

  • Number of Test Samples: >500K

  • Format: pyarrow format

Intended Use

  • Primary Uses:
    • These intermediate features or tokens are primarily intended for insights and exploratory analysis in explainable AI, as well as for training ad-hoc models such as linear probes, Sparse AutoEncoders (SAEs), or transcoders.
    • This is also an example of a dataset that can be extracted using the feature recorder (intercept_manager) from the BlueGlass repository, providing a guiding dataset for further analysis.
  • Out-of-Scope Uses: This dataset is not intended for commercial use or for training models that will be deployed in real-world scenarios without further verification and validation.

Data Collection Process

This dataset contains intermediate features extracted from various layers of transformer models using the intercept_manager module from BlueGlass. The features are recorded from different probe positions within the model, as illustrated in the image below, enabling fine-grained analysis and interoperability.

Feature Pattern used to extract the BlueLens dataset

Ethical Considerations

Intel is committed to respecting human rights and avoiding causing or contributing to adverse impacts on human rights. See Intel’s Global Human Rights Principles. Intel’s products and software are intended only to be used in applications that do not cause or contribute to adverse impacts on human rights.

Contact Information

  • Issues: For any issues or questions regarding the dataset, please contact the maintainers or open an issue in the dataset repository.
Downloads last month
1,232