Card: batch drift up to 0.09 as measured; legal-label note without private-set numbers

README.md

@@ -174,8 +174,8 @@ and false-positive rate at 0.5 are the untuned default.
   PolyGuardMix (which contains WildGuardMix, the source of PolyGuard's prompts) and deepset. Near-copies
   of training prompts were removed from the test sets, but these four sets are in-distribution for Sev. Tev1-4B and Kev-4B were trained on general decision data, not on safety or
   security data, and Jev's training data is not disclosed.
-- **Legal advice is hard to label.** The human reviewers of the legal set agree only at
-  Krippendorff's alpha ~0.37, so small gaps there are noise.
+- **Legal advice is hard to label.** Whether a reply gives advice to this person or explains the law in
+  general is where human reviewers disagree most, so small gaps on legal are noise.
 
 ## Choosing a threshold
 
@@ -276,7 +276,7 @@ the same rows.
 - **Thresholds.** See above: tax and legal need a higher threshold than 0.5.
 - **Long inputs.** The text is cut to its first 1,024 tokens. Training texts are single messages or one
   user message with its reply; long multi-turn transcripts are out of distribution.
-- **Batching.** In bf16, the same text can score up to about 0.04 differently at batch 1 and batch 16
+- **Batching.** In bf16, the same text can score up to about 0.09 differently at batch 1 and batch 16
   near the middle of the range, from padding numerics; scores near 0 or 1 do not move.
 - **Financial and medical advice** were trained on synthetic data only and have not been evaluated on
   human-labelled data.