docs: correct the fertility delta and spell out the eviction example

README.md

@@ -66,13 +66,13 @@ pip install transformers datasets
 python benchmark.py
 ```
 
-It reads a pinned UniMorph commit, five Hugging Face datasets at a fixed document count, and five
-published tokenizers. Nothing from the pipeline that produced this vocabulary is involved.
+It reads a pinned UniMorph commit, five Hugging Face datasets at a fixed document count, and five published tokenizers.
+Nothing from the pipeline that produced this vocabulary is involved.
 
 ### Paradigm sharing
 
-The property this vocabulary is built for. On [UniMorph Hungarian](https://github.com/unimorph/hun)
-(first 20,000 lemmas in file order: 19,424 lemmas, 1,014,882 lemma/form pairs):
+The property this vocabulary is built for.
+On [UniMorph Hungarian](https://github.com/unimorph/hun) first 20,000 lemmas in file order: 19,424 lemmas, 1,014,882 lemma/form pairs:
 
 | tokenizer | vocab | **one stem row + suffixes** | prefix kept | lemma is one token | tokens/form |
 |---|---|---|---|---|---|
@@ -82,13 +82,13 @@ The property this vocabulary is built for. On [UniMorph Hungarian](https://githu
 | FacebookAI/xlm-roberta-base | 250,002 | 10.8% | 60.4% | 12.8% | 3.66 |
 | Qwen/Qwen3-8B | 151,643 | 2.0% | 54.9% | 3.1% | 5.29 |
 
-**one stem row + suffixes** asks both halves of the question at once: the lemma is a SINGLE row,
-*and* its inflected form begins with that row. So every mention of the word teaches one embedding.
+**one stem row + suffixes** asks both halves of the question at once:
+the lemma is a SINGLE row, and its inflected form begins with that row.
+So every mention of the word teaches one embedding.
 
-Neither half can be won on its own. Cutting everything small satisfies `prefix kept` and destroys
-`lemma is one token` -- that is the Qwen row, 54.9% and 2.0%. Swallowing whole inflected forms does
-the reverse. Only the conjunction says a paradigm shares a row, and it does so at the same token
-cost per form as the closest baseline.
+Neither half can be won on its own. Cutting everything small satisfies `prefix kept` and destroys `lemma is one token`.
+Swallowing whole inflected forms does the reverse.
+Only the conjunction says a paradigm shares a row, and it does so at the same token cost per form as the closest baseline.
 
 ### Fertility
 
@@ -102,16 +102,15 @@ Characters a token buys, on the first 2,000 documents of each dataset:
 | `open-web-math/open-web-math` | **2.923** | 2.394 | 2.385 | 3.412 | 3.669 |
 | `bigcode/starcoderdata` (python) | **2.465** | 1.721 | 2.454 | 3.205 | 4.121 |
 
-**On Hungarian prose the two Hungarian baselines beat this tokenizer, and that is the trade, not a
-defect.** PULI buys more characters per token by keeping `házakban` as one row; this vocabulary
-spends an extra token to cut it into `▁ház` `ak` `ban`, and gets the paradigm table above in
-return. A model reading 4% fewer tokens that has to learn `ház` four times over has not come out
-ahead.
+**On Hungarian prose the two Hungarian baselines beat this tokenizer, and that is the trade, not a defect.**
+PULI buys more characters per token by cutting `házakban` into `ház` `akban`, keeping stem and endings fused;
+this vocabulary spends one more token to cut it into `▁ház` `ak` `ban`, and gets the paradigm table above in return.
+A model reading 4% fewer tokens that has to learn `ház` four times over has not come out ahead.
 
-Where the comparison is not about Hungarian morphology it inverts: on mathematics and code this
-vocabulary beats both Hungarian baselines, because a slice of each was in the fit corpus so
-technical text does not fall back to bytes. Qwen wins those outright, as an English-first
-tokenizer should.
+Where the comparison is not about Hungarian morphology it inverts:
+on mathematics and code this vocabulary beats both Hungarian baselines,
+because a slice of each was in the fit corpus so technical text does not fall back to bytes.
+Qwen wins those outright, as an English-first tokenizer should.
 
 ### Special tokens
 
@@ -145,7 +144,8 @@ with a declared Hungarian lexicon deciding what survives:
 1. **Seed** - 32,888 pieces required up front: 28,000 frequent stems, 1,000 named entities,
    1,000 capitalised forms, 432 suffixes, 2,005 alternating stems and 448 programming keywords.
 2. **Overfit** - a large Unigram vocabulary is trained on the corpus.
-3. **Evict** - any piece that is a regular inflection of a declared headword is removed. `okát` is not a row; `ok` and `át` are.
+3. **Evict** - any piece that is a regular inflection of a declared headword is removed. `magyarokat` becomes `▁magyar` `ok` `at`:
+   the whole ending `okat` is not a row, while the plural `ok` and the accusative `at` each are, and every other paradigm reuses them.
 4. **Spare** - an eviction is reversed if the word would then segment worse than its morphology says it should.
 5. **Prune** - rows the corpus never reaches are dropped.