datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
lightonocr-pubtableslightonocr-pubtables-table-onlybankruptcy-early-warning-mlflowfintabnet-hq-tablesmlflow-docsmlflow-standardSamples in this benchmark were generated by RELAI using the following data source(s):
Data Source Name: mlflow
Data Source Link: https://mlflow.org/docs/latest/index.html
Data Source License: https://github.com/mlflow/mlflow/tree/master?tab=Apache-2.0-1-ov-file#readme
Data Source Authors: MLflow Project, a Series of LF Projects, LLC
AI Benchmarks by Data Agents 2025 RELAI.AI. Licensed under CC BY 4.0. Source: https://relai.ai
mlflow-reasoningSamples in this benchmark were generated by RELAI using the following data source(s):
Data Source Name: mlflow
Data Source Link: https://mlflow.org/docs/latest/index.html
Data Source License: https://github.com/mlflow/mlflow/tree/master?tab=Apache-2.0-1-ov-file#readme
Data Source Authors: MLflow Project, a Series of LF Projects, LLC
AI Benchmarks by Data Agents 2025 RELAI.AI. Licensed under CC BY 4.0. Source: https://relai.ai
ain3009-project-mlflow-folderbank-customer-churn-mlflowbank-churn-mlflowmlflow_testpoc-mlflow-tarfile-bypassmlflow-importlib-rce-poc
MLFlow arbitrary code execution via pickle_module_info.txt + loader_module
Vulnerability
CWE-94: Code Injection in MLFlow model loading via importlib.import_module()
on attacker-controlled input.
Three independent attack vectors:
Bug 1: pickle_module_info.txt (mlflow/pytorch/init.py:695)
# Line 815: adds code/ to sys.path
_add_code_from_conf_to_system_path(local_model_path, pytorch_conf)
# Line 683-695: reads pickle_module_info.txt and imports… See the full description on the dataset page: https://huggingface.co/datasets/Talson/mlflow-importlib-rce-poc.distilbert-mlflow-tutoBank-Customer-Churn-MLflow