diff --git a/frameworks/AutoGluon/README.md b/frameworks/AutoGluon/README.md
index 1b5c2dc65..d10bb3a7f 100644
--- a/frameworks/AutoGluon/README.md
+++ b/frameworks/AutoGluon/README.md
@@ -3,3 +3,79 @@
 To run v0.8.2: ```python3 ../automlbenchmark/runbenchmark.py autogluon ...```
 
 To run mainline: ```python3 ../automlbenchmark/runbenchmark.py autogluon:latest ...```
+
+
+## Running with Learning Curves Enabled
+
+To run with learning curves enabled, you must define a new framework in the frameworks YAML file which enables the appropriate parameters. Instructions on how to do this are listed here:
+https://openml.github.io/automlbenchmark/docs/extending/framework/
+
+To summarize these steps:
+1. Add one of the below frameworks (or your own) to [this](https://github.com/openml/automlbenchmark/blob/master/examples/custom/frameworks.yaml) frameworks.yaml file. Alternatively, you can create your own custom user_dir.
+3. Navigate to the root directory of the automlbenchmark repository
+4. Run `python3 runbenchmark.py {YOUR_CUSTOM_FRAMEWORK_NAME} -u {PATH_TO_USER_DIR} ...` \
+    where PATH_TO_USER_DIR is the path to the user_dir containing your frameworks.yaml file. If you used the example user_dir, replace the path with `examples/custom`
+5. For example, if running the most basic framework listed below in the example user_dir, the command would look like: \
+    `python3 runbenchmark.py AutoGluon_curves_true -u examples/custom ...`
+
+### Sample Framework Definitions
+```
+# simplest usage
+AutoGluon_curves_true:
+  extends: AutoGluon
+  params:
+    learning_curves: True
+    _save_artifacts: ['learning_curves']
+```
+```
+# including test data
+AutoGluon_curves_test:
+  extends: AutoGluon
+  params:
+    learning_curves: True
+    _include_test_during_fit: True
+    _save_artifacts: ['learning_curves']
+```
+```
+# parameterizing learning_curves dictionary
+AutoGluon_curves_parameters:
+  extends: AutoGluon
+  params:
+    learning_curves:
+      use_error: True
+    _curve_metrics:
+        binary: ["log_loss", "accuracy", "precision"]
+        regression: ['root_mean_squared_error', 'median_absolute_error', 'r2']
+        multiclass: ["accuracy", "precision_weighted", "recall_weighted"]
+    _include_test_during_fit: True
+    _save_artifacts: ['learning_curves']
+```
+```
+# adding custom hyperparameters
+Defaults: &defaults
+  NN_TORCH:
+    - num_epochs: 1000
+      epochs_wo_improve: 999999999
+  GBM:
+    - num_boost_round: 20000
+      ag_args_fit:
+        early_stop: 999999999
+  XGB:
+    - n_estimators: 20000
+      ag_args_fit:
+        early_stop: 999999999
+
+AutoGluon_curves_hyperparameters:
+  extends: AutoGluon
+  params:
+    hyperparameters:
+      <<: *defaults
+    learning_curves:
+      use_error: True
+    _curve_metrics:
+        binary: ["log_loss", "accuracy", "precision"]
+        regression: ['root_mean_squared_error', 'median_absolute_error', 'r2']
+        multiclass: ["accuracy", "precision_weighted", "recall_weighted"]
+    _include_test_during_fit: True
+    _save_artifacts: ['learning_curves']
+```
\ No newline at end of file
diff --git a/frameworks/AutoGluon/exec.py b/frameworks/AutoGluon/exec.py
index 4b670c4fd..6811f7381 100644
--- a/frameworks/AutoGluon/exec.py
+++ b/frameworks/AutoGluon/exec.py
@@ -16,7 +16,7 @@
 matplotlib.use('agg')  # no need for tk
 
 from autogluon.tabular import TabularPredictor, TabularDataset
-from autogluon.core.utils.savers import save_pd, save_pkl
+from autogluon.core.utils.savers import save_pd, save_pkl, save_json
 import autogluon.core.metrics as metrics
 from autogluon.tabular.version import __version__
 
@@ -68,6 +68,25 @@ def run(dataset, config):
     label = dataset.target.name
     problem_type = dataset.problem_type
 
+    """
+    The _include_test_during_fit flag enables the test_data to be passed into AutoGluon's predictor object
+    during the fit call. If enabled, it is ensured that the test_data is seperated from all training and validation
+    data. It is never seen by the models, nor does it influence the training process in any way.
+
+    One might want to use this flag when generating learning curves. If this flag is enabled and learning_curves
+    have been turned on, then your learning curve artifacts will also include curves for your test dataset.
+    """
+    _include_test_during_fit = config.framework_params.get('_include_test_during_fit', False)
+    if _include_test_during_fit:
+        training_params["test_data"] = test_path
+
+    # whether to generate learning curves (VERY EXPENSIVE. Do not enable for benchmark comparisons.)
+    if "learning_curves" in training_params:
+        lc = training_params["learning_curves"]
+        _curve_metrics = config.framework_params.get('_curve_metrics', {})
+        if isinstance(lc, dict) and "metrics" not in lc and problem_type in _curve_metrics:
+            training_params["learning_curves"]["metrics"] = _curve_metrics[problem_type]
+
     models_dir = tempfile.mkdtemp() + os.sep  # passed to AG
 
     with Timer() as training:
@@ -119,6 +138,7 @@ def inference_time_regression(data: Union[str, pd.DataFrame]):
     prob_labels = probabilities.columns.values.astype(str).tolist() if probabilities is not None else None
     log.info(f"Finished predict in {predict.duration}s.")
 
+    learning_curves = predictor.learning_curves() if training_params.get("learning_curves", None) else None
     _leaderboard_extra_info = config.framework_params.get('_leaderboard_extra_info', False)  # whether to get extra model info (very verbose)
     _leaderboard_test = config.framework_params.get('_leaderboard_test', False)  # whether to compute test scores in leaderboard (expensive)
     leaderboard_kwargs = dict(extra_info=_leaderboard_extra_info)
@@ -136,7 +156,7 @@ def inference_time_regression(data: Union[str, pd.DataFrame]):
     else:
         num_models_ensemble = 1
 
-    save_artifacts(predictor, leaderboard, config)
+    save_artifacts(predictor, leaderboard, learning_curves, config)
     shutil.rmtree(predictor.path, ignore_errors=True)
 
     return result(output_file=config.output_predictions_file,
@@ -151,7 +171,7 @@ def inference_time_regression(data: Union[str, pd.DataFrame]):
                   inference_times=inference_times,)
 
 
-def save_artifacts(predictor, leaderboard, config):
+def save_artifacts(predictor, leaderboard, learning_curves, config):
     artifacts = config.framework_params.get('_save_artifacts', ['leaderboard'])
     try:
         if 'leaderboard' in artifacts:
@@ -167,6 +187,12 @@ def save_artifacts(predictor, leaderboard, config):
             shutil.rmtree(os.path.join(predictor.path, "utils"), ignore_errors=True)
             models_dir = output_subdir("models", config)
             zip_path(predictor.path, os.path.join(models_dir, "models.zip"))
+
+        if 'learning_curves' in artifacts:
+            assert learning_curves is not None, "No learning curves were generated!"
+            learning_curves_dir = output_subdir("learning_curves", config)
+            save_json.save(path=os.path.join(learning_curves_dir, "learning_curves.json"), obj=learning_curves)
+
     except Exception:
         log.warning("Error when saving artifacts.", exc_info=True)