Hugging Face's logo

Datasets:
fenilsonani
/
ZENOBench

Tasks:
Token Classification
Summarization
Languages:
English
Spanish
Tags:
token
llm
ai
machine-learning
agent
code
License:
Dataset card Files
xet
Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot extract the features (columns) for the split 'train' of the config 'default' of the dataset.
Error code:   FeaturesError
Exception:    ArrowInvalid
Message:      Schema at index 1 was different: 
collected_date: string
total_datasets: int64
total_size_bytes: int64
by_source: struct<github: struct<count: int64, size: int64, files: list<item: string>>, jsonplaceholder: struct<count: int64, size: int64, files: list<item: string>>, public_apis: struct<count: int64, size: int64, files: list<item: string>>, synthetic: struct<count: int64, size: int64, files: list<item: string>>>
collected: list<item: struct<source: string, name: string, size: int64, path: string>>
failed: list<item: struct<url: string, error: string>>
vs
sha: string
node_id: string
commit: struct<author: struct<date: string, email: string, name: string>, comment_count: int64, committer: struct<date: string, email: string, name: string>, message: string, tree: struct<sha: string, url: string>, url: string, verification: struct<payload: string, reason: string, signature: string, verified: bool, verified_at: string>>
url: string
html_url: string
comments_url: string
author: struct<avatar_url: string, events_url: string, followers_url: string, following_url: string, gists_url: string, gravatar_id: string, html_url: string, id: int64, login: string, node_id: string, organizations_url: string, received_events_url: string, repos_url: string, site_admin: bool, starred_url: string, subscriptions_url: string, type: string, url: string, user_view_type: string>
committer: struct<avatar_url: string, events_url: string, followers_url: string, following_url: string, gists_url: string, gravatar_id: string, html_url: string, id: int64, login: string, node_id: string, organizations_url: string, received_events_url: string, repos_url: string, site_admin: bool, starred_url: string, subscriptions_url: string, type: string, url: string, user_view_type: string>
parents: list<item: struct<html_url: string, sha: string, url: string>>
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 228, in compute_first_rows_from_streaming_response
                  iterable_dataset = iterable_dataset._resolve_features()
                                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 3496, in _resolve_features
                  features = _infer_features_from_batch(self.with_format(None)._head())
                                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2257, in _head
                  return next(iter(self.iter(batch_size=n)))
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2461, in iter
                  for key, example in iterator:
                                      ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1952, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1974, in _iter_arrow
                  yield from self.ex_iterable._iter_arrow()
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 547, in _iter_arrow
                  yield new_key, pa.Table.from_batches(chunks_buffer)
                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "pyarrow/table.pxi", line 5039, in pyarrow.lib.Table.from_batches
                File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
                File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
              pyarrow.lib.ArrowInvalid: Schema at index 1 was different: 
              collected_date: string
              total_datasets: int64
              total_size_bytes: int64
              by_source: struct<github: struct<count: int64, size: int64, files: list<item: string>>, jsonplaceholder: struct<count: int64, size: int64, files: list<item: string>>, public_apis: struct<count: int64, size: int64, files: list<item: string>>, synthetic: struct<count: int64, size: int64, files: list<item: string>>>
              collected: list<item: struct<source: string, name: string, size: int64, path: string>>
              failed: list<item: struct<url: string, error: string>>
              vs
              sha: string
              node_id: string
              commit: struct<author: struct<date: string, email: string, name: string>, comment_count: int64, committer: struct<date: string, email: string, name: string>, message: string, tree: struct<sha: string, url: string>, url: string, verification: struct<payload: string, reason: string, signature: string, verified: bool, verified_at: string>>
              url: string
              html_url: string
              comments_url: string
              author: struct<avatar_url: string, events_url: string, followers_url: string, following_url: string, gists_url: string, gravatar_id: string, html_url: string, id: int64, login: string, node_id: string, organizations_url: string, received_events_url: string, repos_url: string, site_admin: bool, starred_url: string, subscriptions_url: string, type: string, url: string, user_view_type: string>
              committer: struct<avatar_url: string, events_url: string, followers_url: string, following_url: string, gists_url: string, gravatar_id: string, html_url: string, id: int64, login: string, node_id: string, organizations_url: string, received_events_url: string, repos_url: string, site_admin: bool, starred_url: string, subscriptions_url: string, type: string, url: string, user_view_type: string>
              parents: list<item: struct<html_url: string, sha: string, url: string>>

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

ZENO Benchmark Datasets

This directory contains diverse, real-world JSON datasets collected for benchmarking ZENO's compression performance against JSON and other formats.

Collection Summary

Last Updated: November 5, 2025 Total Datasets: 65 Total Size: 4.17 MB (4,373,804 bytes)

Dataset Categories

1. GitHub API (30 datasets, 1.24 MB)

Real GitHub API responses covering various endpoint types:

Repository Information (10 datasets):

  • Major OSS projects: Kubernetes, React, VS Code, TensorFlow, Rust, Python, Go, Node.js, Django, Vue.js
  • Contains: stars, forks, issues, language stats, license info, etc.
  • Size range: 6-7 KB per file

User Profiles (5 datasets):

  • Famous developers: Linus Torvalds, Guido van Rossum, DHH, Kyle Simpson, TJ Holowaychuk
  • Contains: follower counts, repos, bio, location, etc.
  • Size range: 1-1.5 KB per file

Issues (3 datasets):

  • Open issues from Rust, VS Code, React
  • 10 issues per dataset with full metadata
  • Size range: 54-70 KB per file

Pull Requests (2 datasets):

  • Active PRs from Kubernetes, TensorFlow
  • Includes commits, reviews, labels, assignees
  • Size range: 198-283 KB per file

Contributors (2 datasets):

  • Top 20 contributors from React, Node.js
  • Contains: contributions count, commit stats
  • Size: ~21 KB per file

Commits (2 datasets):

  • Recent 15 commits from Rust, Python
  • Full commit metadata with author, message, stats
  • Size: 77-78 KB per file

Releases (2 datasets):

  • Release history from Rust, Node.js
  • Includes: version tags, release notes, assets
  • Size: 107-217 KB per file

Organizations (4 datasets):

  • Microsoft, Google, Facebook, Apache
  • Org metadata: repos, members, location
  • Size: 1-1.3 KB per file

2. JSONPlaceholder (10 datasets, 235 KB)

Fake REST API data for testing:

  • posts (27 KB): 100 blog posts
  • comments (157 KB): 500 comments
  • albums (9 KB): 100 photo albums
  • photos (10 KB): 50 photo metadata
  • todos (24 KB): 200 todo items
  • users (5 KB): 10 user profiles
  • user_posts (2 KB): Posts by specific user
  • user_albums (1 KB): Albums by specific user
  • post_comments (1 KB): Comments on specific post
  • posts_single (292 bytes): Single post detail

3. Public APIs (10 datasets, 2.34 MB)

Data from various public REST APIs:

Geographic Data:

  • countries_usa (21 KB): USA country information
  • countries_region_europe (263 KB): All European countries
  • countries_language_spanish (122 KB): Spanish-speaking countries

Cryptocurrency Data:

  • crypto_coins_list (1.6 MB): Complete cryptocurrency list (15,000+ coins)
  • crypto_bitcoin (140 KB): Bitcoin detailed data
  • crypto_ethereum (140 KB): Ethereum detailed data
  • crypto_markets (47 KB): Top 50 crypto markets

Other APIs:

  • dog_breeds (4 KB): Dog breed catalog
  • breweries (25 KB): 50 brewery records
  • nested_posts_with_comments (10 KB): Posts with embedded comments

4. Synthetic Datasets (15 datasets, 463 KB)

Carefully crafted datasets representing common real-world patterns:

E-commerce & Business:

  • ecommerce_catalog (31 KB): 100 products with varied attributes
  • user_profiles (34 KB): 80 user accounts with preferences
  • npm_packages (40 KB): 60 package.json configurations

Logging & Events:

  • server_logs (48 KB): 200 structured log entries
  • event_stream (41 KB): 150 event records
  • api_responses (51 KB): 50 API response samples

Time-series & Sensor Data:

  • sensor_timeseries (32 KB): 150 sensor readings
  • geographic_data (12 KB): 60 city records with coordinates

Compression Test Cases:

  • numeric_sequences (1.9 KB): Linear, Fibonacci, powers, primes (delta compression test)
  • repeated_values (11 KB): High repetition data (sparse mode test)
  • wide_table (21 KB): 50 records × 20 fields (column mode test)
  • database_records_sparse (13 KB): Records with sparse fields

Complex Structures:

  • nested_structures (22 KB): Deeply nested objects
  • mixed_types (23 KB): All JSON types mixed
  • large_text_fields (86 KB): Articles with large text content

Size Distribution

Size Range Count Example
< 10 KB 28 User profiles, organizations, small configs
10-50 KB 24 Repository info, logs, synthetic data
50-100 KB 7 Issues, commits, large text fields
100-300 KB 7 Releases, crypto data, country lists
> 300 KB 1 Crypto coins list (1.6 MB)

Diversity Characteristics

The dataset collection includes:

  • Multiple domains: Development tools, social media, e-commerce, finance, geography
  • Various structures: Flat objects, nested hierarchies, arrays, mixed types
  • Different patterns:
    • Repetitive data (sparse mode candidates)
    • Numeric sequences (delta encoding candidates)
    • Wide tables (column mode candidates)
    • String-heavy content (dictionary candidates)
  • Size variety: 292 bytes to 1.6 MB
  • Real-world APIs: GitHub, CoinGecko, REST Countries, etc.
  • Representative synthetic data: Common use cases like logs, configs, time-series

File Structure

Each dataset consists of two files:

source/
  dataset_name.json           # The actual JSON data
  dataset_name.metadata.json  # Collection metadata

Metadata Format 

{
  "id": "source_dataset_name",
  "source": "github|jsonplaceholder|public_apis|synthetic",
  "collected_date": "2025-11-05T22:51:39.293685",
  "size_bytes": 12345,
  "description": "Brief description of the dataset",
  "url": "Original URL if applicable"
}

Usage in Benchmarks

These datasets are used to:

  1. Measure compression ratios: ZENO vs JSON vs MessagePack vs Protobuf
  2. Test encoding modes: Verify column/row/sparse/delta selection
  3. Benchmark performance: Encoding/decoding speed across dataset types
  4. Validate correctness: Round-trip testing (JSON → ZENO → JSON)
  5. Token efficiency: Estimate LLM token reduction

Expanding the Dataset

To scale to 1000+ datasets:

  1. GitHub API: Expand to more repos, longer histories
  2. Package Registries: npm, PyPI, crates.io package metadata
  3. Social Media: Twitter/X, Reddit public APIs
  4. Open Data: Government datasets, scientific data
  5. Logs: Real production logs (anonymized)
  6. Configurations: Real-world config files from OSS projects
  7. Database Dumps: Sample SQL→JSON exports

License & Attribution

  • GitHub API data: Public API, subject to GitHub's terms
  • JSONPlaceholder: Fake data, free to use
  • Public APIs: Various licenses, check individual sources
  • Synthetic data: Generated for this project, no restrictions

All data is for benchmarking and research purposes only

Downloads last month
256