Dataset Viewer
Auto-converted to Parquet
doc
stringlengths
14
66
split
stringclasses
3 values
domain
stringclasses
53 values
10 : 1 = 100 : 10
train
math_division10
10 : 1 = 30 : 3
train
math_division10
10 : 1 = 40 : 4
train
math_division10
10 : 1 = 60 : 6
train
math_division10
10 : 1 = 70 : 7
train
math_division10
10 : 100 = 12 : 144
train
math_squares
10 : 100 = 14 : 196
train
math_squares
10 : 100 = 16 : 256
train
math_squares
10 : 100 = 18 : 324
train
math_squares
10 : 100 = 2 : 4
train
math_squares
10 : 100 = 22 : 484
train
math_squares
10 : 100 = 25 : 625
train
math_squares
10 : 100 = 8 : 64
train
math_squares
10 : 2 = 15 : 3
train
math_division5
10 : 2 = 20 : 4
train
math_division5
10 : 2 = 30 : 6
train
math_division5
10 : 2 = 35 : 7
train
math_division5
10 : 2 = 5 : 1
train
math_division5
10 : 2 = 50 : 10
train
math_division5
10 : 2 = 55 : 11
train
math_division5
10 : 2 = 70 : 14
train
math_division5
10 : 2 = 85 : 17
train
math_division5
10 : 20 = 100 : 200
train
math_double
10 : 20 = 12 : 24
train
math_double
10 : 20 = 14 : 28
train
math_double
10 : 20 = 15 : 30
train
math_double
10 : 20 = 2 : 4
train
math_double
10 : 20 = 22 : 44
train
math_double
10 : 20 = 24 : 48
train
math_double
10 : 20 = 28 : 56
train
math_double
10 : 20 = 30 : 60
train
math_double
10 : 20 = 32 : 64
train
math_double
10 : 20 = 34 : 68
train
math_double
10 : 20 = 35 : 70
train
math_double
10 : 20 = 38 : 76
train
math_double
10 : 20 = 40 : 80
train
math_double
10 : 20 = 45 : 90
train
math_double
10 : 20 = 46 : 92
train
math_double
10 : 20 = 5 : 10
train
math_double
10 : 20 = 50 : 100
train
math_double
10 : 20 = 54 : 108
train
math_double
10 : 20 = 55 : 110
train
math_double
10 : 20 = 62 : 124
train
math_double
10 : 20 = 72 : 144
train
math_double
10 : 20 = 74 : 148
train
math_double
10 : 20 = 76 : 152
train
math_double
10 : 20 = 78 : 156
train
math_double
10 : 20 = 82 : 164
train
math_double
10 : 20 = 84 : 168
train
math_double
10 : 20 = 86 : 172
train
math_double
10 : 20 = 90 : 180
train
math_double
10 : 20 = 94 : 188
train
math_double
10 : 20 = 96 : 192
train
math_double
10 : 20 = 98 : 196
train
math_double
10 : 5 = 12 : 6
train
math_division2
10 : 5 = 14 : 7
train
math_division2
10 : 5 = 2 : 1
train
math_division2
10 : 5 = 20 : 10
train
math_division2
10 : 5 = 22 : 11
train
math_division2
10 : 5 = 28 : 14
train
math_division2
10 : 5 = 34 : 17
train
math_division2
10 : 5 = 42 : 21
train
math_division2
10 : 5 = 48 : 24
train
math_division2
10 : 5 = 54 : 27
train
math_division2
10 : 5 = 56 : 28
train
math_division2
10 : 5 = 6 : 3
train
math_division2
10 : 5 = 60 : 30
train
math_division2
10 : 5 = 62 : 31
train
math_division2
10 : 5 = 66 : 33
train
math_division2
10 : 5 = 68 : 34
train
math_division2
10 : 5 = 70 : 35
train
math_division2
10 : 5 = 74 : 37
train
math_division2
10 : 5 = 8 : 4
train
math_division2
10 : 5 = 84 : 42
train
math_division2
10 : 5 = 86 : 43
train
math_division2
10 : 5 = 88 : 44
train
math_division2
10 : 5 = 92 : 46
train
math_division2
10 : 5 = 94 : 47
train
math_division2
10 : 5 = 96 : 48
train
math_division2
100 : 10 = 10 : 1
train
math_division10
100 : 10 = 16 : 4
train
math_root
100 : 10 = 30 : 3
train
math_division10
100 : 10 = 36 : 6
train
math_root
100 : 10 = 40 : 4
train
math_division10
100 : 10 = 60 : 6
train
math_division10
100 : 10 = 70 : 7
train
math_division10
100 : 20 = 15 : 3
train
math_division5
100 : 20 = 20 : 4
train
math_division5
100 : 20 = 30 : 6
train
math_division5
100 : 20 = 35 : 7
train
math_division5
100 : 20 = 5 : 1
train
math_division5
100 : 20 = 50 : 10
train
math_division5
100 : 20 = 55 : 11
train
math_division5
100 : 20 = 70 : 14
train
math_division5
100 : 20 = 85 : 17
train
math_division5
100 : 200 = 12 : 24
train
math_double
100 : 200 = 14 : 28
train
math_double
100 : 200 = 15 : 30
train
math_double
100 : 200 = 2 : 4
train
math_double
100 : 200 = 22 : 44
train
math_double
End of preview. Expand in Data Studio

Dataset Card for hyperdimensional probe

This repository contains the official datasets of "Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures".

  • This work combines symbolic representations and neural probing to introduce Hyperdimensional Probe, a new paradigm for decoding LLM vector space into human-interpretable features, consistently extracting meaningful concepts across models and inputs.

Datasets

  1. Corpus of factual and linguistic analogies (input-completition tasks): saturnMars/hyperprobe-dataset-analogy
  2. SQuAD-based corpus (question-answering tasks): saturnMars/hyperprobe-dataset-squad

Dataset Details

Dataset Description

This repository includes our syntethic corpora for the training and experimental stages.

Information

Dataset Sources

Examples

Train data

[
  " 10 : 1 = 60 : 6",
  " 10 : 100 = 12 : 144",
  " plato : kepler = philosopher : mathematician",
  " significant : successful = significantly : successfully",
  " important : importantly = subsequent : subsequently"],
  " 10 : 100 = 28 : 784",
  " coyote : canine = cat : feline",
  " coyote : canine = cow : bovid",
  " sold : oversold = played : overplayed",
  " sold : oversold = populated : overpopulated"],
  " 10 : 1 = 80 : 8",
  " rarely : quietly = rare : quiet",
  " rarely : rare = calmly : calm",
  " rarely : rare = critically : critical",
  " youngest : young = sweetest : sweet"]
]

Test data

{
    "capital_world": [
        " Athens is to Greece as Baghdad is to Iraq",
        " Athens is to Greece as Bangkok is to Thailand"],
    "currency": [
        " Algeria is to dinar as Angola is to kwanza",
        " Algeria is to dinar as Brazil is to real"],
    "family": [
        " boy is to girl as brother is to sister",
        " boy is to girl as dad is to mom"],
    "comparative": [
        " bad is to worse as big is to bigger",
        " bad is to worse as bright is to brighter"],
    "verb_Ving_3pSg": [
        " adding is to adds as advertising is to advertises",
        " adding is to adds as appearing is to appears"],
    "male_female": [
        " actor is to actress as batman is to batwoman",
        " actor is to actress as boy is to girl"]
}

Source Data

This corpora were generated using two knowledge bases:

  1. Google Analogy Test Set;
  2. The Bigger Analogy Test Set (BATS).

Google Analogy Test Set is distributed by TensorFlow under the Apache License 2.0, whereas BATS is released under the CC-BY-NC 4.0 License.

  • See the GitHub repository to reconstruct the corpora from the these two knowledge bases.

Citation

If you use any of these datasets in your research, please cite the following work:

@misc{bronzini2025hyperdimensional,
    title={Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures},
    author={Marco Bronzini and Carlo Nicolini and Bruno Lepri and Jacopo Staiano and Andrea Passerini},
    year={2025},
    eprint={2509.25045},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

APA: Bronzini, M., Nicolini, C., Lepri, B., Staiano, J., & Passerini, A. (2025). Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures.

Downloads last month
34