dataset card with multi-config + verdict stats

README.md

@@ -45,6 +45,10 @@ ds_verbose = load_dataset("massaindustries/dataset-A-routing", name="verbose", s
 
 # Prompts + ground truth (no model outputs) — per replicare i test
 ds_evals = load_dataset("massaindustries/dataset-A-routing", name="evals", split="train")
+
+# Tutti i config hanno una riga `_schema_anchor` (query_id == '_schema_anchor') con valori
+# dummy per fissare lo schema di datasets/PyArrow. Filtrala via:
+#   ds = ds.filter(lambda r: r["query_id"] != "_schema_anchor")
 ```
 
 ## Schema `results`
@@ -95,9 +99,9 @@ Tutti i campi di `results` + per ogni modello m in {qwen, ds4, kimi}:
 
 | modello | correct | incorrect | abstention | accuracy |
 |---|---|---|---|---|
-| qwen | 3477 | 1786 | 241 | 0.6606 |
-| ds4 | 4056 | 1220 | 228 | 0.7688 |
-| kimi | 4129 | 784 | 591 | 0.8404 |
+| qwen | 3478 | 1786 | 241 | 0.6607 |
+| ds4 | 4057 | 1220 | 228 | 0.7688 |
+| kimi | 4130 | 784 | 591 | 0.8405 |
 
 ## Coverage check