spaicom-lab/semasia-cifar100

Latents for cifar100 (timm)

   

This repository hosts precomputed latent representations (embeddings) extracted from timm image-classification backbones on cifar100, released as part of SEMASIA — a large-scale resource for studying semantic communication, cross-model latent space alignment, and explainability. Each config corresponds to a single model; only that model's Parquet files are read on load_dataset.

Usage

Load with datasets and convert to torch:

from datasets import load_dataset
import torch

ds = load_dataset(
    "spaicom-lab/semasia-cifar100",  # repository  →  which benchmark
    "aimv2_1b_patch14_224.apple_pt",  # config      →  which model
    split="test",  # split       →  which partition
).with_format("torch")

embeddings = torch.vstack(list(ds["embedding"]))  # (N, d)
fine_label = torch.tensor(ds["fine_label"])  # (N,)

Or read the Parquet files directly with polars:

import polars as pl

df = pl.read_parquet(
    "hf://datasets/spaicom-lab/semasia-cifar100/test/aimv2_1b_patch14_224.apple_pt/*.parquet"
)

embeddings = df["embedding"].to_numpy()  # shape (N, d)
fine_label = df["fine_label"].to_numpy()  # shape (N,)

Fields

Columns available in each Parquet file for this dataset:

Field	Description
id	Row index within the shard (unique per split/model, not across models); the row order matches the original uoft-cs/cifar100 split, so id can be used to map a row back to its source sample.
model_name	timm model that produced this row's embedding (constant within a config).
embedding	Precomputed latent representation extracted by the model (dimensionality depends on config; see model registry).
fine_label	Original dataset field, copied as-is from the source dataset.
coarse_label	Original dataset field, copied as-is from the source dataset.

Available Models

Number of models with precomputed embeddings, per split:

Split	# Models
test	1699
train	1699

Notes

• Configs are generated from what is actually uploaded on the Hub (parquet presence).
• Based on uoft-cs/cifar100
• Code: github.com/SPAICOM/semasia-datasets
• Model metadata (architecture family, parameter count, embedding dimension, pretraining details, ...) for every model in this dataset is available in the model registry.

Citation

If you use this dataset, please cite:

@misc{pandolfo2026semasialargescaledatasetsemantically,
      title={SEMASIA: A Large-Scale Dataset of Semantically Structured Latent Representations},
      author={Mario Edoardo Pandolfo and Enrico Grimaldi and Lorenzo Marinucci and Leonardo Di Nino and Simone Fiorellino and Sergio Barbarossa and Paolo Di Lorenzo},
      year={2026},
      eprint={2605.09485},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2605.09485},
}