Fix math numbers (0.468/0.208 vs 0.456/0.212)

README.md

@@ -21,10 +21,12 @@ measure the pedagogy judge's noise floor against training noise.
 |---|---|---|
 | repo | [`socrateach-sft-olmo2-1b-lora`](https://huggingface.co/meric533/socrateach-sft-olmo2-1b-lora) | `socrateach-sft-olmo2-1b-lora-run2` |
 | judge OVERALL (no SI, deduped) | 0.685 | 0.527 |
-| GSM8K hinted | 0.556 | 0.552 |
+| GSM8K bare | 0.468 | 0.456 |
+| GSM8K hinted | 0.208 | 0.212 |
 
-Same data, same hyperparameters, same code. They differ only in run-to-run nondeterminism,
-and they are genuinely different weights — not a re-upload of one checkpoint.
+Same recipe, independently trained (run 1 comes from the earlier POC). They are genuinely
+different weights, not a re-upload of one checkpoint. Note how tightly the *math* numbers
+agree — 0.012 and 0.004 apart — compared to the judge gap below.
 
 ## Why this pair matters