Dataset Viewer
Auto-converted to Parquet Duplicate
generated_at
string
test_pairs
int64
positive_pairs
int64
negative_pairs
int64
doc_hit_rate
float64
keyword_hit_rate
float64
negative_low_confidence_rate
float64
avg_confidence
float64
avg_confidence_all
float64
per_document
dict
thresholds
dict
passed
bool
2026-07-29T08:41:53.429601+00:00
21
18
3
1
1
0.6667
0.4284
0.3967
{ "hr_leave_policy_en": 1, "hr_leave_policy_fa": 1, "hse_safety_guidelines_en": 1, "procurement_process_en": 1, "it_acceptable_use_en": 1, "remote_work_policy_en": 1 }
{ "min_doc_hit_rate": 0.85, "min_keyword_hit_rate": 0.75, "min_avg_confidence": 0.4, "max_negative_confidence": 0.35, "min_negative_low_confidence_rate": 0.66 }
true

OrgMind Enterprise Policy Samples

Synthetic organizational policy documents and QA benchmark pairs for OrgMind RAG Studio.

Contents

File Description
chunks.jsonl Semantic chunks with page/paragraph citation metadata
qa_pairs.jsonl Curated questions with expected document + keywords
benchmark_report.json Reproducible retrieval metrics
eval_results.json Benchmark summary (no per-row details)
manifest.json Corpus statistics

Documents (synthetic)

  • HR Leave Policy (EN + FA)
  • HSE Safety Guidelines
  • Procurement Process
  • IT Acceptable Use
  • Remote Work Policy

All content is fictional and created for demonstration purposes.

Reproduce

python scripts/build_dataset.py
python scripts/train_index.py
python scripts/benchmark.py
Downloads last month
42