Expired50% confidenceBenchmarkExact time
Qwen 3.6 Max in neutral mode has an ambiguity detection F1 of only 16% and asks an average of only 0.07 questions per task, but once it does ask for clarification, the accuracy reaches 94.7%
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/9/2026
First Seen
Valid until: 10/7/2026(expired)
Sources
Related Claims
UnverifiedMiniMax M2.7 asks 0.6 questions per task, but its clarification advancement rate is only 60.7%, indicating a problem of low-information-gain questioning71% similarUnverifiedQwen 3.6 Max has an end-to-end accuracy of only 12.3% in neutral mode69% similarUnverifiedQwen3.7 Max的输出质量高度依赖提示词质量，同一模型在不同提示词下的性能差异可高达30%-50%68% similarUnverifiedIn guided mode, the average end-to-end accuracy increased from 28.6% to 33.7%, and the ambiguity detection F1 score increased from 45.3% to 64.9%68% similarUnverifiedThe Qwen3.6 27B version achieved an accuracy of 95.7% on the Simple QA test.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/387230API
curl https://kongchang.com/api/v1/knowledge/claims/387230MCP
get_claim(id=387230)