Revert suffix stripping from arXiv search query

Searching the full model name (e.g. "Mistral 7B") is more precise —
papers that use "Mistral" generically may refer to different variants.
Also removes the now-unused re import and restores the original test.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: erwardenaar

src/openllm_selector/database.py

@@ -2,7 +2,6 @@
 
 import json
 import pathlib
-import re
 import xml.etree.ElementTree as ET
 
 import requests
@@ -509,14 +508,10 @@ def fetch_recent_papers(model_name: str, max_results: int = 3) -> list[dict]:
     >>> papers[0].keys()
     dict_keys(['title', 'authors', 'summary', 'published', 'arxiv_url'])
     """
-    # Strip trailing size/context suffixes (e.g. "7B", "32B", "1.4B", "4K") so
-    # "OLMo 2 32B" searches as "OLMo 2". Mixture-of-experts names like "8x7B"
-    # are kept because they are part of the canonical model identity.
-    search_name = re.sub(r"\s+\d+(\.\d+)?[BbMmKk]$", "", model_name)
     response = requests.get(
         "https://export.arxiv.org/api/query",
         params={
-            "search_query": f'all:"{search_name}"',
+            "search_query": f'all:"{model_name}"',
             "sortBy": "submittedDate",
             "sortOrder": "descending",
             "max_results": max_results,

tests/test_database.py

@@ -1617,9 +1617,7 @@ def test_model_name_used_in_query(self):
             mock_get.return_value = _mock_response(_SAMPLE_ATOM)
             fetch_recent_papers("Pythia 6.9B")
         _, kwargs = mock_get.call_args
-        # Size suffix is stripped before searching; "Pythia 6.9B" -> "Pythia"
-        assert "Pythia" in kwargs["params"]["search_query"]
-        assert "6.9B" not in kwargs["params"]["search_query"]
+        assert "Pythia 6.9B" in kwargs["params"]["search_query"]
 
     def test_empty_feed_returns_empty_list(self):
         with patch("openllm_selector.database.requests.get") as mock_get: