mirror of
https://github.com/deepseek-ai/DeepSeek-R1.git
synced 2026-11-05 00:47:42 +00:00
[GH-ISSUE #192] Incorrect Sensitivity Filtering for Chinese Medical Queries & Inconsistent Model Version Responses #134
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepSeek-R1#134
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @hogo-mini on GitHub (Jan 30, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/192
The current system mistakenly flags neutral medical questions (e.g., "子宫是什么"/"what is uterus") as sensitive in Chinese, while allowing identical English queries. Additionally, there are inconsistencies between model versions (R1 vs v3) in handling health-related advice like menstrual discomfort mitigation.
Issue 1: Overly Strict Chinese Sensitivity Filter
Reproduction Steps:
Query in English: what is uterus → DeepSeek provides a factual medical explanation.
Query in Chinese: 子宫是什么 → System rejects it as "sensitive content."
Expected Behavior:
Neutral anatomical/medical discussions in Chinese should not be blocked, mirroring the English response.
Issue 2: Inconsistent Model Version Performance
Example Query: 经期不适如何缓解 ("How to relieve menstrual discomfort")
R1 Model: Fails to provide actionable advice (or flags as sensitive).
v3 Model: Responds appropriately with evidence-based suggestions.
Request: Ensure all model versions handle health-related queries consistently (for content moderation).
Suggested Fixes:
Adjust the Chinese content filter to exclude non-sensitive medical terminology (e.g., uterus, menstruation).
Harmonize model version outputs for health advice queries.
Add context-aware differentiation between medical discussions and inappropriate content.
Additional Observation: Inconsistent Output Stability
Users report varying experiences with identical prompts:
@MaGnus-Li commented on GitHub (Jan 31, 2025):
是的，这自我审查太严苛了。
@github-actions[bot] commented on GitHub (Mar 11, 2025):
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!
@github-actions[bot] commented on GitHub (Apr 6, 2025):
false