[
    {
        "model_name": "nicl",
        "model_type": "Tabular Foundation Model",
        "model_description": "Our in-house tabular foundation model based on an in-context learning architecture."
    },
  {
    "model_name": "xgboost",
    "model_type": "Decision Trees",
    "model_description": "A highly efficient and scalable gradient boosting framework that improves performance with regularization and handles large tabular datasets effectively."
  },
  {
    "model_name": "xgboost_ensemble",
    "model_type": "Decision Trees",
    "model_description": "(Ensembled Version) A highly efficient and scalable gradient boosting framework that improves performance with regularization and handles large tabular datasets effectively."
  },
  {
    "model_name": "catboost",
    "model_type": "Decision Trees",
    "model_description": "A high-performance gradient boosting algorithm that handles categorical features natively and requires minimal preprocessing."
  },
  {
    "model_name": "catboost_ensemble",
    "model_type": "Decision Trees",
    "model_description": "(Ensembled Version) A high-performance gradient boosting algorithm that handles categorical features natively and requires minimal preprocessing."
  },
  {
    "model_name": "modern-nca",
    "model_type": "Neural Networks",
    "model_description": "An advanced extension of Neighborhood Component Analysis (NCA) that employs deep neural networks with stochastic neighborhood sampling to learn complex feature interactions for tabular data."
  },
  {
    "model_name": "tabpfn2.5",
    "model_type": "Tabular Foundation Model",
    "model_description": "A transformer-based model that performs in-context learning by approximating Bayesian inference for tabular classification on small datasets."
  },
  {
    "model_name": "realmlp",
    "model_type": "Neural Networks",
    "model_description": "An enhanced MLP optimized with strong default parameters, achieving competitive performance on tabular data. Note that due to computate limitation, we used the simple version of realmlp."
  },
  {
    "model_name": "tabicl",
    "model_type": "Tabular Foundation Model",
    "model_description": "A transformer-based model that performs feature compression before doing in-context learning on tabular data by conditioning on labeled support examples to predict unseen queries without task-specific training."
  },
  {
    "model_name": "lightgbm",
    "model_type": "Decision Trees",
    "model_description": "A fast gradient boosting framework that builds decision trees using histogram-based learning for scalable, high-performance tabular modeling."
  },
  {
    "model_name": "lightgbm_ensemble",
    "model_type": "Decision Trees",
    "model_description": "(Ensembled Version) A fast gradient boosting framework that builds decision trees using histogram-based learning for scalable, high-performance tabular modeling."
  },
  {
    "model_name": "tabm",
    "model_type": "Neural Networks",
    "model_description": "A deep learning model that efficiently imitates an Neural Networks, enhancing performance on tabular data tasks without the overhead of training multiple models."
  }
]