GPT-OSS 120B × debug

**Bug #1 – Filtering out words that occur only once**  

```python
return [word for word, count in ranked[:n] if count > 1]
```

*What it does wrong*  
The list‑comprehension keeps only those words whose count is **greater than 1**.  
When the requested `n` is larger than the number of words that appear more than once, the
function returns fewer than `n` items and, more importantly, it discards perfectly valid
“single‑occurrence” words.  
In the first test case (`'the cat sat on the mat the'`) the words *cat* and *mat* appear
once, yet they are required in the output.

*Fix*  

```python
return [word for word, _ in ranked[:n]]
```

We simply return the first `n` items from the sorted list, regardless of their frequency
(the underscore is used because the count is no longer needed).

---

**Bug #2 – Empty strings are counted as words**  

```python
words = [w.strip('.,!?;:\'"') for w in words]
```

*What it does wrong*  
If a token consists solely of punctuation (e.g. `"!!!"` or `"..."`) the `strip` call turns it
into an empty string `''`. The empty string is then handed to `Counter`, which treats it as
a legitimate word and can end up in the final result (e.g. `['', …]`).  

*Fix*  

```python
words = [w.strip(string.punctuation) for w in words if w.strip(string.punctuation)]
```

We (a) strip **all** punctuation characters (see Bug #3) and (b) keep only those tokens that
are non‑empty after stripping.

---

**Bug #3 – Incomplete punctuation set**  

```python
w.strip('.,!?;:\'"')
```

*What it does wrong*  
The hard‑coded string of punctuation characters misses many symbols that appear in real
text (parentheses, brackets, hyphens, back‑ticks, etc.). Consequently those characters stay
attached to words and change their spelling (`"hello-"` → `"hello-"` instead of `"hello"`).

*Fix*  

```python
import string
...
words = [w.strip(string.punctuation) for w in words if w.strip(string.punctuation)]
```

`string.punctuation` contains the full set of ASCII punctuation characters, guaranteeing
consistent stripping.

---

### Corrected function

```python
import string
from collections import Counter
from typing import List

def top_words(text: str, n: int = 10) -> List[str]:
    """
    Return the `n` most frequent words in *text*.
    Words are compared case‑insensitively, sorted by descending frequency,
    and alphabetically for ties.
    """
    # 1. split on whitespace and normalise case
    raw_words = text.lower().split()

    # 2. strip *all* leading/trailing punctuation and discard empty results
    words = [
        w.strip(string.punctuation)
        for w in raw_words
        if w.strip(string.punctuation)
    ]

    # 3. count occurrences
    counts = Counter(words)

    # 4. sort by (-frequency, word) → frequency descending, then alphabetically
    ranked = sorted(counts.items(), key=lambda item: (-item[1], item[0]))

    # 5. return the top‑`n` words (no frequency filter)
    return [word for word, _ in ranked[:n]]
```

The function now passes all the supplied test cases:

```python
>>> top_words('the cat sat on the mat the', 3)
['the', 'cat', 'mat']

>>> top_words('a a b b c', 2)
['a', 'b']

>>> top_words('hello world', 5)
['hello', 'world']
```

All three bugs have been identified, explained, and fixed.