diff --git a/clip_benchmark/cli.py b/clip_benchmark/cli.py
index d9434d7..df75327 100644
--- a/clip_benchmark/cli.py
+++ b/clip_benchmark/cli.py
@@ -59,7 +59,7 @@ def get_parser_args():
     parser_eval.add_argument('--dump_templates', default=False, action="store_true", help="dump templates to the results json file.")
 
     parser_eval.add_argument('--language', default="en", type=str, nargs="+", help="language(s) of classname and prompts to use for zeroshot classification.")
-    parser_eval.add_argument('--output', default="result.json", type=str, help="output file where to dump the metrics. Can be in form of a template, e.g., --output='{dataset}_{pretrained}_{model}_{language}_{task}.json'")
+    parser_eval.add_argument('--output', default="{dataset}_{pretrained}_{model}_{language}_{task}.json", type=str, help="output file where to dump the metrics. Can be in form of a template, e.g., --output='{dataset}_{pretrained}_{model}_{language}_{task}.json'")
     parser_eval.add_argument('--quiet', dest='verbose', action="store_false", help="suppress verbose messages")
     parser_eval.add_argument('--save_clf', default=None, type=str, help="optionally save the classification layer output by the text tower")
     parser_eval.add_argument('--load_clfs', nargs='+', default=[], type=str, help="optionally load and average mutliple layers output by text towers.")