diff --git a/branch-config/filelist b/branch-config/filelist
index 03e1bc0..0ba2d96 100644
--- a/branch-config/filelist
+++ b/branch-config/filelist
@@ -3,3 +3,10 @@ main.py
 requirements.txt
 evaluation_examples
 logs
+
+mm_agents
+experiment_a11y_tree.py
+experiment_screenshot.py
+experiment_screenshot_a11y_tree.py
+experiment_screenshot_seeact.py
+experiment_screenshot_som.py
diff --git a/branch_flag b/branch_flag
index 9daeafb..180b47c 100644
--- a/branch_flag
+++ b/branch_flag
@@ -1 +1 @@
-test
+baseline
diff --git a/desktop_env/envs/desktop_env.py b/desktop_env/envs/desktop_env.py
index ef7a1ad..f8cd082 100644
--- a/desktop_env/envs/desktop_env.py
+++ b/desktop_env/envs/desktop_env.py
@@ -76,6 +76,8 @@ class DesktopEnv(gym.Env):
         self.cache_dir_base: str = cache_dir
         self.vm_screen_size = screen_size
 
+        os.makedirs(self.tmp_dir_base, exist_ok=True)
+
         # task-aware stuffs
         # todo: handling the logic of snapshot directory
         self._set_task_info(task_config)
diff --git a/desktop_env/evaluators/metrics/table.py b/desktop_env/evaluators/metrics/table.py
index 1bbdc60..d7e4faf 100644
--- a/desktop_env/evaluators/metrics/table.py
+++ b/desktop_env/evaluators/metrics/table.py
@@ -92,8 +92,11 @@ def compare_table(result: str, expected: str, **options) -> float:
     if result is None:
         return 0.
 
-    xlworkbookr: Workbook = openpyxl.load_workbook(filename=result)
-    pdworkbookr = pd.ExcelFile(result)
+    try:
+        xlworkbookr: Workbook = openpyxl.load_workbook(filename=result)
+        pdworkbookr = pd.ExcelFile(result)
+    except:
+        return 0.
     worksheetr_names: List[str] = pdworkbookr.sheet_names
 
     xlworkbooke: Workbook = openpyxl.load_workbook(filename=expected)
diff --git a/desktop_env/evaluators/metrics/utils.py b/desktop_env/evaluators/metrics/utils.py
index 08a3ca8..4f35b65 100644
--- a/desktop_env/evaluators/metrics/utils.py
+++ b/desktop_env/evaluators/metrics/utils.py
@@ -52,14 +52,17 @@ def load_sparklines(xlsx_file: str, sheet_name: str) -> Dict[str, str]:
     """
 
     # read xlsx
-    with zipfile.ZipFile(xlsx_file, "r") as z_f:
-        with z_f.open("xl/workbook.xml") as f:
-            workbook_database: _Element = lxml.etree.fromstring(f.read())
-            sheets: List[_Element] = _sheet_name_selector(workbook_database)
-            sheet_names: Dict[str, str] = {sh.get("name"): sh.get("sheetId") for sh in sheets}
-        with z_f.open("xl/worksheets/sheet{:}.xml".format(sheet_names[sheet_name])) as f:
-            sheet: _Element = lxml.etree.fromstring(f.read())
-            sparklines: List[_Element] = _sparklines_selector(sheet)
+    try:
+        with zipfile.ZipFile(xlsx_file, "r") as z_f:
+            with z_f.open("xl/workbook.xml") as f:
+                workbook_database: _Element = lxml.etree.fromstring(f.read())
+                sheets: List[_Element] = _sheet_name_selector(workbook_database)
+                sheet_names: Dict[str, str] = {sh.get("name"): sh.get("sheetId") for sh in sheets}
+            with z_f.open("xl/worksheets/sheet{:}.xml".format(sheet_names[sheet_name])) as f:
+                sheet: _Element = lxml.etree.fromstring(f.read())
+                sparklines: List[_Element] = _sparklines_selector(sheet)
+    except zipfile.BadZipFile:
+        return {}
 
     sparklines_dict: Dict[str, str] = {}
     for sp_l in sparklines:
@@ -158,29 +161,32 @@ def load_charts(xlsx_file: Workbook, sheet_name: str, **options) -> Dict[str, An
 _shared_str_selector = lxml.cssselect.CSSSelector("oo|sst>oo|si>oo|t", namespaces=_xlsx_ns_mapping)
 def read_cell_value(xlsx_file: str, sheet_name: str, coordinate: str) -> Any:
     #  read_cell_value {{{ # 
-    with zipfile.ZipFile(xlsx_file, "r") as z_f:
-        try:
-            with z_f.open("xl/sharedStrings.xml") as f:
-                shared_str_xml: _Element = lxml.etree.fromstring(f.read())
-                str_elements: List[_Element] = _shared_str_selector(shared_str_xml)
-                shared_strs: List[str] = [elm.text for elm in str_elements]
-        except:
-            logger.debug("Read shared strings error: %s", xlsx_file)
+    try:
+        with zipfile.ZipFile(xlsx_file, "r") as z_f:
+            try:
+                with z_f.open("xl/sharedStrings.xml") as f:
+                    shared_str_xml: _Element = lxml.etree.fromstring(f.read())
+                    str_elements: List[_Element] = _shared_str_selector(shared_str_xml)
+                    shared_strs: List[str] = [elm.text for elm in str_elements]
+            except:
+                logger.debug("Read shared strings error: %s", xlsx_file)
 
-        with z_f.open("xl/workbook.xml") as f:
-            workbook_database: _Element = lxml.etree.fromstring(f.read())
-            sheets: List[_Element] = _sheet_name_selector(workbook_database)
-            sheet_names: Dict[str, str] = {sh.get("name"): sh.get("sheetId") for sh in sheets}
+            with z_f.open("xl/workbook.xml") as f:
+                workbook_database: _Element = lxml.etree.fromstring(f.read())
+                sheets: List[_Element] = _sheet_name_selector(workbook_database)
+                sheet_names: Dict[str, str] = {sh.get("name"): sh.get("sheetId") for sh in sheets}
 
-        with z_f.open("xl/worksheets/sheet{:}.xml".format(sheet_names[sheet_name])) as f:
-            sheet: _Element = lxml.etree.fromstring(f.read())
-            cells: List[_Element] =\
-                    lxml.cssselect.CSSSelector( 'oo|row>oo|c[r="{:}"]'.format(coordinate)
-                                              , namespaces=_xlsx_ns_mapping
-                                              )(sheet)
-            if len(cells)==0:
-                return None
-            cell: _Element = cells[0]
+            with z_f.open("xl/worksheets/sheet{:}.xml".format(sheet_names[sheet_name])) as f:
+                sheet: _Element = lxml.etree.fromstring(f.read())
+                cells: List[_Element] =\
+                        lxml.cssselect.CSSSelector( 'oo|row>oo|c[r="{:}"]'.format(coordinate)
+                                                  , namespaces=_xlsx_ns_mapping
+                                                  )(sheet)
+                if len(cells)==0:
+                    return None
+                cell: _Element = cells[0]
+    except zipfile.BadZipFile:
+        return None
 
     cell: Dict[str, str] = xmltodict.parse( lxml.etree.tostring(cell, encoding="unicode")
                                           , process_namespaces=True
diff --git a/experiment_screenshot_a11y_tree.py b/experiment_screenshot_a11y_tree.py
index 60c81b6..c48a2d7 100644
--- a/experiment_screenshot_a11y_tree.py
+++ b/experiment_screenshot_a11y_tree.py
@@ -41,7 +41,8 @@ logger.addHandler(sdebug_handler)
 
 logger = logging.getLogger("desktopenv.experiment")
 
-PATH_TO_VM = r"C:\Users\tianbaox\Documents\Virtual Machines\Ubuntu\Ubuntu.vmx"
+#PATH_TO_VM = r"C:\Users\tianbaox\Documents\Virtual Machines\Ubuntu\Ubuntu.vmx"
+PATH_TO_VM = "../../../../大文件/镜像/Ubuntu-1218/Ubuntu/Ubuntu.vmx"
 
 
 def run_one_example(example, agent, max_steps=10, example_trajectory_dir="exp_trajectory", recording=True):
@@ -109,8 +110,9 @@ def run_one_example(example, agent, max_steps=10, example_trajectory_dir="exp_tr
 
 if __name__ == "__main__":
     action_space = "pyautogui"
-    example_class = "chrome"
-    example_id = "7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3"
+    example_class = "libreoffice_calc"
+    #example_id = "7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3"
+    example_id = "01b269ae-2111-4a07-81fd-3fcd711993b0"
     gpt4_model = "gpt-4-vision-preview"
     gemini_model = "gemini-pro-vision"
 
@@ -120,7 +122,8 @@ if __name__ == "__main__":
 
     with open(f"evaluation_examples/examples/{example_class}/{example_id}.json", "r") as f:
         example = json.load(f)
-    example["snapshot"] = "exp_setup4"
+    #example["snapshot"] = "exp_setup4"
+    example["snapshot"] = "Snapshot 30"
 
     api_key = os.environ.get("OPENAI_API_KEY")
     agent = GPT4v_Agent(api_key=api_key, model=gpt4_model, instruction=example['instruction'],
diff --git a/main.py b/main.py
index 2027736..a3a01df 100644
--- a/main.py
+++ b/main.py
@@ -46,9 +46,9 @@ def human_agent():
     Runs the Gym environment with human input.
     """
 
-    with open("evaluation_examples/examples/libreoffice_calc/a01fbce3-2793-461f-ab86-43680ccbae25.json.nosetup", "r") as f:
+    with open("evaluation_examples/examples/thunderbird/7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3.json", "r") as f:
         example = json.load(f)
-    example["snapshot"] = "Snapshot 26"
+    example["snapshot"] = "Snapshot 30"
 
     env = DesktopEnv( path_to_vm="../../../../大文件/镜像/Ubuntu-1218/Ubuntu/Ubuntu.vmx"
                     , action_space="computer_13"
diff --git a/mm_agents/gpt_4v_agent.py b/mm_agents/gpt_4v_agent.py
index cc08b79..55fd2a5 100644
--- a/mm_agents/gpt_4v_agent.py
+++ b/mm_agents/gpt_4v_agent.py
@@ -22,6 +22,10 @@ from mm_agents.prompts import SYS_PROMPT_IN_SCREENSHOT_OUT_CODE, SYS_PROMPT_IN_S
     SYS_PROMPT_IN_SOM_A11Y_OUT_TAG, \
     SYS_PROMPT_SEEACT, ACTION_DESCRIPTION_PROMPT_SEEACT, ACTION_GROUNDING_PROMPT_SEEACT
 
+import logging
+
+logger = logging.getLogger("desktopenv.agent")
+
 
 # Function to encode the image
 def encode_image(image_path):
@@ -247,6 +251,7 @@ class GPT4v_Agent:
             if self.exp == "both":
                 _screenshot = previous_obs["screenshot"]
                 _linearized_accessibility_tree = previous_obs["accessibility_tree"]
+                logger.debug("LINEAR AT: %s", _linearized_accessibility_tree)
 
                 messages.append({
                     "role": "user",
@@ -451,7 +456,7 @@ class GPT4v_Agent:
             "max_tokens": self.max_tokens
         })
 
-        print(response)
+        logger.debug("RESPONSE: %s", response)
 
         if self.exp == "seeact":
             messages.append({
diff --git a/requirements.txt b/requirements.txt
index 68ca347..b494102 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -35,3 +35,6 @@ playwright
 backoff
 formulas
 pydrive
+fastdtw
+
+openai