Skip to content

update error judge #393

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
e06084 merged 1 commit into from
Apr 17, 2026
+5 −1
Merged

update error judge #393

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion dingo/model/llm/llm_document_parsing_ocr.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,12 +103,16 @@ def process_response(cls, response: str) -> EvalDetail:
json_match = re.search(r'\{[\s\S]*"errors"[\s\S]*\}', response)
types = []
names = []
parse_ok = False
errors_nonempty = False

if json_match:
try:
json_str = json_match.group()
result_data = json.loads(json_str)
errors = result_data.get("errors", [])
parse_ok = True
errors_nonempty = len(errors) > 0
Comment on lines 113 to +115
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Apr 17, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The current implementation is not robust against unexpected JSON structures from the LLM. If the errors field is not a list (e.g., a string or null), len(errors) will raise a TypeError, and the subsequent loop will also fail. It's safer to verify that errors is indeed a list before proceeding.

Suggested change
errors = result_data.get("errors", [])
parse_ok = True
errors_nonempty = len(errors) > 0
errors = result_data.get("errors", [])
if isinstance(errors, list):
parse_ok = True
errors_nonempty = bool(errors)
else:
errors = []


for error in errors:
error_category = error.get("error_category", "")
Expand All @@ -123,7 +127,7 @@ def process_response(cls, response: str) -> EvalDetail:
log.error("未找到JSON内容")

result = EvalDetail(metric=cls.__name__)
result.status = False
result.status = (not parse_ok) or errors_nonempty
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Apr 17, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The updated logic for result.status correctly identifies issues, but the subsequent code for constructing result.label has two problems:

  1. If no errors are found (status is False), the label becomes ["."] because types and names are empty. It should ideally be ["QUALITY_GOOD"].
  2. If parsing fails (parse_ok is False), the label also becomes ["."], which is not descriptive.

Returning early for the success case and setting a specific label for parsing errors would improve the output quality.

Suggested change
result.status = (not parse_ok) or errors_nonempty
result.status = (not parse_ok) or errors_nonempty
if not result.status:
result.label = ["QUALITY_GOOD"]
result.reason = [json_str] if 'json_str' in locals() else [response]
return result
if not parse_ok:
types, names = ["QUALITY_BAD"], ["ParseError"]


tmp_type = '.'.join(types)
tmp_name = '.'.join(names)
Expand Down
Loading