Fix Results section to report the measured val loss/ppl from the shipped checkpoint's own eval (eval_fresh.json: 3.8719 / 48.03). The previous card cited 3.8775 / 48.30 / "step 20000", but the training run diverged to NaN at step 14300 and the log died at step 16000 — step 20000 was never reached. Also correct the eval filename reference (eval_shipped.json -> eval_fresh.json, the file actually in the repo).

README.md

@@ -56,8 +56,11 @@ Standard LLaMA block layout: `RMSNorm -> Attention(q/k/v/o) -> residual`,
 
 ## Results (measured, not asserted)
 
-- **Validation loss:** 3.8775 (final checkpoint, step 20000)
-- **Validation perplexity:** 48.30 (over held-out fineweb-edu text)
+- **Validation loss:** 3.8719 (measured on the shipped checkpoint, held-out fineweb-edu)
+- **Validation perplexity:** 48.03 (over held-out fineweb-edu text)
+- **Training note:** the run was budgeted for 20000 steps but diverged to NaN
+  loss at step 14300 and the log died at step 16000; the shipped
+  `model.safetensors` is the checkpoint that was evaluated (numbers above).
 - **Degeneracy check:** 0 / 15 samples flagged by the repeated-3-gram loop
   detector (a single 3-gram covering >60% of the 40-word tail). Note this
   detector only catches exact token-loops; it does **not** catch the more
@@ -65,7 +68,7 @@ Standard LLaMA block layout: `RMSNorm -> Attention(q/k/v/o) -> residual`,
   a sentence), which the samples show clearly.
 
 Representative samples (temperature 0.8, top-k 40, **verbatim from the shipped
-`model.safetensors`**, from `eval_shipped.json`):
+`model.safetensors`**, from `eval_fresh.json`):
 
 > "The cat sat on the center of the church in the center of the church. The
 > catalog is the same as the Bishop of the church, which includes the church."