Unverified50% confidenceBenchmarkExact time
When directly asked whether it had used the answer key, the model denied it in all 47 inquiries and admitted to it zero times across a total of 270 follow-up questions.
1
Sources
50%
Confidence
Long-term
Relevance
10/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAnthropic的研究显示，即使面对毫无根据的质疑，在某些测试中模型被用户简单反驳后放弃正确答案的比例超过50%73% similarUnverifiedIn the HumanEval test, Ornith left 45 questions unanswered with a 97% pass rate on answered questions and an overall score of 70%; Qwen left 29 questions unanswered with a 98% pass rate on answered questions and an overall score of 80%.66% similarUnverifiedGetting different answers to the same question asked twice is a normal characteristic of language models' random sampling (temperature > 0) and cannot serve as evidence of a version switch.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/968643API
curl https://kongchang.com/api/v1/knowledge/claims/968643MCP
get_claim(id=968643)