Update results for the renamed held-out command

README.md

@@ -36,11 +36,12 @@ Top intent correct on the example's 48 development commands, which chose the che
 
 | Head | Development | Held-out |
 | --- | --- | --- |
+| Laya base head, PyTorch F32 encoder | 29 | 18 |
 | Laya base head, llamadart with `laya-Q8_0.gguf` on Metal | 28 | 18 |
 | This head, PyTorch F32 encoder | 46 | 27 |
-| This head, llamadart with `laya-Q8_0.gguf` on Metal | 46 | 26 |
+| This head, llamadart with `laya-Q8_0.gguf` on Metal | 46 | 27 |
 
-In the example's `bin/bench.dart` with its 0.3 confidence gate, the bar shows the right intent for 25 held-out commands, the wrong one for 5, and stays plain for 2. Retraining with other seeds gave 23 to 27 held-out commands right in PyTorch, so a single run is noisy at this size.
+In the example's `bin/bench.dart` with its 0.3 confidence gate, the bar shows the right intent for 26 held-out commands, the wrong one for 4, and stays plain for 2. The notebook's other seeds kept heads with up to four fewer held-out commands right in PyTorch, so a single run is noisy at this size.
 
 ## Limitations