Note 15M was the Ultra-MC smoke test

BENCHMARKS.md

@@ -89,6 +89,7 @@ val_ppl 59.8 at 419M tokens. No full `run_eval.py` on 68k yet.
 
 ## 114M sibling
 
+This 15M run was the toy / smoke-test for
 [Byrne-100M-Ultra-MC](https://huggingface.co/Quazim0t0/Byrne-100M-Ultra-MC)
-is the larger looped run (Memory Cache, fractal RoPE, ~2.0B tokens).
-Different size, data, objectives. Not a matched scale-up of this 15M.
+(Memory Cache, fractal RoPE, ~2.0B tokens). Different size, data,
+objectives. Not a matched scale-up.