diff --git a/happypose/pose_estimators/megapose/src/megapose/evaluation/bop.py b/happypose/pose_estimators/megapose/src/megapose/evaluation/bop.py
index 43c8a1c9..96a25390 100644
--- a/happypose/pose_estimators/megapose/src/megapose/evaluation/bop.py
+++ b/happypose/pose_estimators/megapose/src/megapose/evaluation/bop.py
@@ -182,9 +182,9 @@ def run_evaluation(cfg: BOPEvalConfig) -> None:
     results_path = Path(cfg.results_path)
     eval_dir = Path(cfg.eval_dir)
 
-    if cfg.dataset == "hb" and cfg.split == "test":
+    if cfg.dataset == "hb.bop19" and cfg.split == "test":
         cfg.convert_only = True
-    if cfg.dataset == "itodd" and cfg.split == "test":
+    if cfg.dataset == "itodd.bop19" and cfg.split == "test":
         cfg.convert_only = True
 
     scores_pose_path = None
@@ -196,7 +196,8 @@ def run_evaluation(cfg: BOPEvalConfig) -> None:
         # this is what is expected by bop_toolkit_lib
         csv_path = eval_dir / f"{method}_{cfg.dataset.split('.')[0]}-{cfg.split}.csv"
 
-        convert_results_to_bop(results_path, csv_path, cfg.method, use_pose_score=False)
+        # pose scores give better AR scores in general
+        convert_results_to_bop(results_path, csv_path, cfg.method, use_pose_score=True)
 
         if not cfg.convert_only:
             _run_bop_evaluation(csv_path, cfg.eval_dir, eval_detection=False)