patch: minor bug fixes for evaluator and task configurations, documentation update (#121)

* fix: /cursor_position api return format fix

* chore: update README.md to remove deprecated command

* fix: add base score for evaluators and minor bug fixes

* fix: add base score for setup configurations

---------

Co-authored-by: Jiaqi Deng <jiaqideng@Jiaqis-MacBook-Pro.local>

desktop_env/evaluators/metrics/docs.py

@@ -47,7 +47,7 @@ def find_default_font(config_file_path, rules):
     return 1 if default_font == expected_font else 0
 
 
-def contains_page_break(docx_file):
+def contains_page_break(docx_file, rules):
     if not docx_file:
         return 0
 
@@ -57,6 +57,11 @@ def contains_page_break(docx_file):
         logger.error(f"Error: {e}")
         return 0
 
+    try:
+        expected_page_break_count = rules["page_break_count"]
+    except Exception as e:
+        expected_page_break_count = None
+
     namespaces = {'w': 'http://schemas.openxmlformats.org/wordprocessingml/2006/main'}
 
     for paragraph in doc.paragraphs:
@@ -65,9 +70,15 @@ def contains_page_break(docx_file):
             for br in br_elems:
                 if br is not None and '{http://schemas.openxmlformats.org/wordprocessingml/2006/main}type' in br.attrib and \
                         br.attrib['{http://schemas.openxmlformats.org/wordprocessingml/2006/main}type'] == 'page':
-                    return 1
-    return 0
+                    page_break_count += 1
+    
+    if expected_page_break_count is not None and page_break_count != expected_page_break_count:
+        return 0    
 
+    if page_break_count > 0:
+        return 1
+    else:
+        return 0
 
 def compare_docx_files(file1, file2, **options):
     ignore_blanks = options.get('ignore_blanks', True)
@@ -867,7 +878,10 @@ def compare_references(file1, file2, **options):
         total_similarity += similarity
 
     result = total_similarity / len(ref1)
-    if result >= reference_base_result:
+
+    epsilon = 0.01
+
+    if result >= reference_base_result + epsilon:
         return (result - reference_base_result) / (1 - reference_base_result)
     else:
         return 0