Demote the proper-noun anecdote out of the result tables

bench3.1/README.md

@@ -415,9 +415,14 @@ None of which tells you what it *sounds* like. A native speaker might well hear
 
 ---
 
-## One word, one language
+## An anecdote, and why it is not a result
 
-The French prompt says *"la lumière de **Marseille**"*. Six phonemes out of seventy — the error rate drowns them, the ear does not:
+**This measures nothing.** One word, in one language out of eleven, with no denominator — it is recorded
+here because it was noticed, not because it supports a conclusion. It is deliberately absent from every
+table above.
+
+The French prompt contains a city name. The four 4B files render it as a non-word on almost every seed,
+the 8B files render it correctly on all three:
 
 | | seed 42 | seed 100 k | seed 100 M |
 |---|---|---|---|
@@ -427,7 +432,13 @@ The French prompt says *"la lumière de **Marseille**"*. Six phonemes out of sev
 | `4b-v3.1-mlp` | ✗ | ✗ | ✓ |
 | `4b-v3.1` | ✗ | ✗ | ✗ |
 
-Treat this for what it is: **one proper noun, in one of eleven languages, on three seeds.** It is not a ranking criterion and the aggregate scores already say 4B and 8B are equivalent. It is a hint that the two sizes may diverge on rare lexical items even where they agree on everything else — consistent with quantisation costing facts, which was already known. If your prompts lean on proper nouns, test both before deciding.
+What keeps it from being a finding, beyond the sample size: **it costs almost nothing on the sentence.**
+`4b-v3.1` mangles the name outright and ends up **one** phoneme worse than `8b-v3.1` over the whole line.
+Every aggregate metric in this report is blind to it, which cuts both ways — they cannot confirm it either.
+
+The aggregate scores say 4B and 8B are equivalent, and that is the conclusion to keep. The only reason to
+mention this at all is that quantisation is independently known to cost factual recall: if your prompts
+lean on names of people or places, test both sizes on **your** prompts rather than trusting anything here.
 
 ---
 
@@ -437,10 +448,11 @@ Treat this for what it is: **one proper noun, in one of eleven languages, on thr
 |---|---|
 | generate images or video without speech | **`4b-ClipProj-v3.1`** — 4.6 GB, indistinguishable from the 32B |
 | are tight on VRAM | **`4b-ClipProj-v3.1`** — the ridge is 26 MB and gives up nothing measurable |
-| generate multilingual speech | **`8b-ClipProj-v3.1`** — best speech score, and steady on proper nouns |
-| rely on named people or places | **`8b-ClipProj-v3.1`**, and run the 32B once to check the name works there at all |
+| generate multilingual speech | **`4b-ClipProj-v3.1`** covers nine of the eleven languages tested; `8b-ClipProj-v3.1` has the best overall speech score, by less than the seed-to-seed spread |
+| have the VRAM to spare | `8b-ClipProj-v3.1` — nothing measured says you need it, nothing says it hurts |
 
-Do not take a v3: it is the only difference this benchmark resolves cleanly.
+**Do not take a v3.** That is the only difference this benchmark resolves cleanly: v3 versus v3.1 is real,
+4B versus 8B is not, ridge versus residual is not.
 
 ---