diff --git a/sam_audio/model/model.py b/sam_audio/model/model.py
index 1bf5c586..2ac5ddc7 100644
--- a/sam_audio/model/model.py
+++ b/sam_audio/model/model.py
@@ -267,6 +267,18 @@ def separate(
                 ),
             )
 
+            # Refresh anchor conditioning created by predict_spans()
+            forward_args.update(
+                {
+                    "anchor_ids": self._repeat_for_reranking(
+                        batch.anchor_ids, reranking_candidates
+                    ),
+                    "anchor_alignment": self._repeat_for_reranking(
+                        batch.anchor_alignment, reranking_candidates
+                    ),
+                }
+            )
+
         audio_features = forward_args["audio_features"]
         B, T, C = audio_features.shape
         C = C // 2  # we stack audio_features, so the actual channels is half