Understand the project
Practical overview
When it fits
Use Cleanlab after an initial model has produced prediction probabilities or feature embeddings and the team wants to investigate dataset quality before another training iteration.
Who it is for
- Machine-learning teams validating training data
- Data scientists working with noisy or real-world labels
- Teams managing image, text, audio or tabular datasets
- Projects using multiple annotators or active-learning workflows
Documented uses
- Find likely label errors in a dataset
- Identify outliers, duplicates and other data issues
- Profile a dataset before or during model development
- Prioritize records for labeling or relabeling
- Support data-centric model improvement without changing the modeling framework
Before you choose
Requirements & limits
The repository is public, open source under the Apache-2.0 license, and documents installation through uv, pip or conda. Python 3.10 or later is required; Linux, macOS and Windows are supported.
- The workflow depends on an existing model's prediction probabilities or feature embeddings.
- The library covers several modalities and task types rather than only tabular classification.
- Cleanlab can be used with models from multiple major ML frameworks and libraries.
- The package supports analysis of multi-annotator data and active-learning prioritization.
- Teams should plan a human or operational review stage for issues identified by the analysis.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Evidence and transparency
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 2 October 2026
- Original website title
- cleanlab/cleanlab · Official homepage title · retained in its source language