Packages

crucible_datasets

0.1.0
0.5.4 0.5.3 0.5.2 0.5.1 0.5.0 0.4.1 retired 0.4.0 retired 0.2.0 0.1.0

Dataset management and caching for AI research benchmarks

HexDocs

Current section

Files

Jump to
Readme
9 Versions
2 Dependencies
Dependants
Files
Activity
crucible_datasets CHANGELOG.md
Raw

CHANGELOG.md 

# Changelog



All notable changes to this project will be documented in this file.



## [0.1.0] - 2025-10-07



### Added

- Initial release

- Centralized dataset management for AI evaluation research

- Unified interface for benchmark datasets (MMLU, HumanEval, GSM8K)

- Automatic caching with version tracking for fast access

- Comprehensive evaluation metrics (exact match, F1 score, custom metrics)

- Dataset sampling strategies (random, stratified, k-fold cross-validation)

- Support for custom datasets from local JSONL files

- Reproducibility features with deterministic sampling and version control



### Documentation

- Comprehensive README with examples

- API documentation for dataset loading and evaluation

- Usage examples for common research workflows

- Integration guide for research infrastructure libraries