Team Ai
Datasetpublic

Humanbased-AI/Crypto-Address-Annotation-10K

Codatta Crypto Address Annotations (Sample) Overview This dataset is a 10,000-row sample of the comprehensive Codatta Crypto Address Annotations database. The full database serves as a massive repository of over 500 million labeled address pairs across multiple blockchains. The data provides critical metadata aimed at solving the problem of fragmented and siloed blockchain information. It includes entity names, functional categories (e.g., Exchanges, DeFi, Scam)… See the full description on the dataset page: https://huggingface.co/datasets/Humanbased-AI/Crypto-Address-Annotation-10K.

sourceHugging Faceopenrailupdated 10mo agoView on Hugging Face
1likes237downloads
Dataset Card

Codatta Crypto Address Annotations (Sample)

Overview

This dataset is a 10,000-row sample of the comprehensive Codatta Crypto Address Annotations database. The full database serves as a massive repository of over 500 million labeled address pairs across multiple blockchains.

The data provides critical metadata aimed at solving the problem of fragmented and siloed blockchain information. It includes entity names, functional categories (e.g., Exchanges, DeFi, Scam), and risk assessments. The dataset is curated through a unique hybrid approach that bridges crowdsourced intelligence with standardized open protocols:

  • —Codatta Platform: Aggregates data from a decentralized data intelligence network, verified by a community of contributors.
  • —Microscope Protocol: Utilizes an open-source initiative for collaborative crypto address labeling to ensure high fidelity and standardization.

Dataset Contents

The dataset contains structured metadata for crypto addresses. The specific data fields included in this sample are:

  • —`chain` (string): The blockchain network (e.g., bitcoin, ethereum).
  • —`address` (string): The specific wallet or smart contract address.
  • —`name` (string): The specific label name (e.g., "Binance Deposit Address").
  • —`entity` (string): The entity that owns or controls the address (e.g., "Binance").
  • —`category` (string): Functional classification (e.g., CEX, DeFi, Scam).
  • —Reference: Full category list available at Microscope Protocol Categories.
  • —`source` (string): The origin of the label, such as ground_truth, external, heuristics, or machine_learning.

Key Statistics

While this repository provides a sample for testing, the complete Codatta database covers a massive scale of on-chain entities. Below is the distribution of the full 517.61 Million labeled pairs.

1. Distribution by Chain

(Total labeled pairs: 517,609,906)

ChainAddress Count
Bitcoin185,818,811
Polygon129,251,435
Ethereum107,420,551
BSC (BNB Chain)61,043,196
Tron22,055,631
Optimism4,926,779
Arbitrum3,488,054
Avalanche3,369,625
Base233,495

2. Distribution by Category (Top Categories)

CategoryAddress Count
Smart Contract235,023,302
Exchange187,716,225
Scam24,261,860
Gambling21,677,685
Ransom17,991,759
DeFi User16,825,002
Bridging User11,516,923
DEX User7,286,609
ERC20 Token5,442,026
Mixer5,186,762
Wallet2,724,811
FIAT2,145,036
Darknet2,110,659
Sanctioned1,101,861

Usage

This 10k sample is released publicly to foster research, demonstrate data quality, and allow developers to test data integration.

  • —Research & Testing: Use this sample to evaluate the schema, label accuracy, and coverage for tasks like token classification or entity recognition.
  • —Commercial & Advanced Use: Users interested in accessing the full dataset (500M+ labels) for commercial applications, advanced analytics, or security integration should contact the Codatta team directly.

License and Open-Source Details