From 115373ab304828c683bbb2b38f1102a3e0c5a197 Mon Sep 17 00:00:00 2001
From: "jinli.yl" <jinli.yl@alibaba-inc.com>
Date: Thu, 26 Feb 2026 18:20:12 +0800
Subject: [PATCH] refactor(benchmark): remove unused llm config and update test
 parameters

---
 benchmark/halumem/eval_reme.py                        | 11 -----------
 pyproject.toml                                        |  6 ++++++
 .../vector_based/personal/personal_retriever.py       |  6 +++---
 tests/test_chunking_utils.py                          |  7 ++-----
 4 files changed, 11 insertions(+), 19 deletions(-)

diff --git a/benchmark/halumem/eval_reme.py b/benchmark/halumem/eval_reme.py
index 229cfb85..9c71478a 100644
--- a/benchmark/halumem/eval_reme.py
+++ b/benchmark/halumem/eval_reme.py
@@ -320,7 +320,6 @@ class MemoryProcessor:
                 return_dict=True,
                 enable_time_filter=True,
                 enable_thinking_params=self.enable_thinking_params,
-                llm_config_name="qwen-plus-t",
             )
 
             duration_ms = (time.time() - start) * 1000
@@ -355,7 +354,6 @@ class MemoryProcessor:
             return_dict=True,
             enable_time_filter=True,
             enable_thinking_params=self.enable_thinking_params,
-            llm_config_name="qwen-plus-t",
         )
 
         # Extract memories from response
@@ -569,15 +567,6 @@ class HaluMemEvaluator:
             default_llm_config={
                 "model_name": self.config.reme_model_name,
             },
-            llms={
-                "qwen-plus-t": {
-                    "backend": "openai",
-                    "model_name": "qwen-plus",
-                    "extra_body": {
-                        "enable_thinking": True,
-                    },
-                },
-            },
         )
 
         # Load evaluation prompts into ReMe's prompt handler
diff --git a/pyproject.toml b/pyproject.toml
index 028f4fa8..19df32d9 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -113,5 +113,11 @@ remecli = "reme.reme_cli:main"
 
 [tool.pytest.ini_options]
 asyncio_default_fixture_loop_scope = "function"
+# Script-style tests that need to be run with `python test_*.py`
+testpaths = ["tests"]
+python_files = ["test_*.py"]
+python_functions = ["test_*"]
+# Exclude script-style tests that require manual execution
+addopts = "--ignore=tests/test_embedding.py --ignore=tests/test_embedding_cache.py --ignore=tests/test_embedding_sync.py --ignore=tests/test_file_store.py"
 
 # python -m build && twine upload dist/*
diff --git a/reme/memory/vector_based/personal/personal_retriever.py b/reme/memory/vector_based/personal/personal_retriever.py
index 517cddd2..b8803717 100644
--- a/reme/memory/vector_based/personal/personal_retriever.py
+++ b/reme/memory/vector_based/personal/personal_retriever.py
@@ -12,9 +12,9 @@ class PersonalRetriever(BaseMemoryAgent):
 
     clear && python benchmark/halumem/eval_reme.py \
         --data_path /Users/yuli/workspace/HaluMem/data/HaluMem-Medium.jsonl \
-        --reme_model_name qwen3-30b-a3b-instruct-2507 \
-        --algo_version v1 \
-        --enable_thinking_params
+        --reme_model_name qwen3.5-plus \
+        --batch_size 10000 \
+        --algo_version default
 
     📊 Question Answering (with LLM answer):
       Correct (all):       0.8537
diff --git a/tests/test_chunking_utils.py b/tests/test_chunking_utils.py
index c0937021..0ac44d0a 100644
--- a/tests/test_chunking_utils.py
+++ b/tests/test_chunking_utils.py
@@ -42,11 +42,8 @@ def test_chunk_markdown_empty():
         overlap=10,
     )
 
-    # Empty string splits to [""] which creates one chunk with empty text
-    assert len(chunks) == 1
-    assert chunks[0].text == ""
-    assert chunks[0].start_line == 1
-    assert chunks[0].end_line == 1
+    # Empty text is filtered out (no meaningful content to store)
+    assert len(chunks) == 0
 
 
 def test_chunk_markdown_single_line():