hardness
Datasets
All datasets matching “hardness”hardness_data_mix
Hardness Data Mix - Resolution Sufficiency Dataset
A large-scale dataset of document images with labels indicating the minimum resolution required to accurately answer questions about those documents.
Dataset Description
This dataset contains 81,924 document image-question pairs labeled with resolution sufficiency information. Each sample is annotated with a "hardness" label indicating the minimum resolution level needed to answer questions about that document accurately.… See the full description on the dataset page: https://huggingface.co/datasets/Kimhi/hardness_data_mix.AM11_DED316L_texture_hardness_agent_only
Dataset Card
Dataset Description
[Placeholder]
Dataset Structure
[Placeholder]
Uses
[Placeholder]
Limitations
[Placeholder]
License
[Placeholder]
Citation
[Placeholder]
AM11_DED316L_texture_hardness_agent_only_deprecated
Dataset Card
Dataset Description
[Placeholder]
Dataset Structure
[Placeholder]
Uses
[Placeholder]
Limitations
[Placeholder]
License
[Placeholder]
Citation
[Placeholder]
levir-oacp-r2-hardness-adaptive-seed42-runsus-water-hardness
US Water Hardness Dataset
Measured drinking-water hardness for 16,561 US cities — mg/L as calcium carbonate and grains per gallon (gpg), each value carrying its source, tier, and sample date.
Canonical, continuously-updated source: https://www.tapwaterdata.com/water-hardness
GitHub mirror: https://github.com/Tapwaterdata/water-hardness-dataset
DOI: 10.5281/zenodo.21315903
Files
water-hardness.csv — 16,561 rows, 17-column header.
water-hardness.json — same 17… See the full description on the dataset page: https://huggingface.co/datasets/tapwaterdata/us-water-hardness.audioset24k-mae-hardness
AudioSet-Opus MAE reconstruction hardness
Per-clip MAE reconstruction loss over
danjacobellis/audioset_opus_24kbps,
computed with a frozen dasheng-base encoder + trained MAE decoder. Loss is the
per-patch normalised reconstruction MSE on masked, in-span patches, averaged over
4 fixed-seed masking passes. Higher loss = harder example.
Columns: id, path, loss, loss_all_patches, n_valid_patches,
duration_s, labels, rank (1 = hardest).