CoolFace
Datasetpublic

latkes/factprobe-replication-stage1-cooccurrence-v1

factprobe-replication-stage1-cooccurrence-v1 How many documents of the OLMo-2 pretraining corpus contain BOTH names of a probed pair. Computed over all 1,117 token files (15.50 TB, 3.875 trillion tokens) for the 2,172,383 name pairs the model was probed about; 313,576 of them share at least one document. Replaces an earlier version measured on the 192 GB mid-training mix alone, where 80-92% of pairs never co-occurred and the quantity behaved as a yes/no flag rather than a graded… See the full description on the dataset page: https://huggingface.co/datasets/latkes/factprobe-replication-stage1-cooccurrence-v1.

sourceHugging Facemitupdated 1mo agoView on Hugging Face
0likes13downloads
Dataset Card

factprobe-replication-stage1-cooccurrence-v1

How many documents of the OLMo-2 pretraining corpus contain BOTH names of a probed pair. Computed over all 1,117 token files (15.50 TB, 3.875 trillion tokens) for the 2,172,383 name pairs the model was probed about; 313,576 of them share at least one document. Replaces an earlier version measured on the 192 GB mid-training mix alone, where 80-92% of pairs never co-occurred and the quantity behaved as a yes/no flag rather than a graded measure.

Dataset Info

  • —Rows: 313576
  • —Columns: 3

Columns

ColumnTypeDescription
aValue('string')first name of the pair, sorted so a pair is always written the same way
bValue('string')second name of the pair
documentsValue('int64')how many documents contain both names

Generation Parameters

json
{
  "script_name": "cooccurrence_shard.py",
  "model": "OLMo-2 pretraining corpus (not a model run)",
  "description": "How many documents of the OLMo-2 pretraining corpus contain BOTH names of a probed pair. Computed over all 1,117 token files (15.50 TB, 3.875 trillion tokens) for the 2,172,383 name pairs the model was probed about; 313,576 of them share at least one document. Replaces an earlier version measured on the 192 GB mid-training mix alone, where 80-92% of pairs never co-occurred and the quantity behaved as a yes/no flag rather than a graded measure.",
  "experiment_name": "factprobe-replication",
  "job_id": "mll:72741",
  "cluster": "mll",
  "artifact_status": "final",
  "canary": false,
  "hyperparameters": {
    "source": "document index written during counting, 437 GB over 1,117 shards",
    "name_matching": "OLMo token sequences, AI2 splitting rule, word boundaries both ends, both written forms",
    "pairs_considered": 2172383
  },
  "input_datasets": [
    "olmo-mix-1124 stage-1 pretraining corpus"
  ]
}

Usage

python
from datasets import load_dataset

dataset = load_dataset("latkes/factprobe-replication-stage1-cooccurrence-v1", split="train")
print(f"Loaded {len(dataset)} rows")