Unverified50% confidenceBenchmarkExact time
A single "trusted source" prompt caused 7 out of 8 tested models to reverse between 45% and 88% of their correct answers.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/1/2026
First Seen
Valid until: 12/30/2026
Sources
Related Entities
Related Claims
UnverifiedThe 2023 TruthfulQA benchmark showed that even the strongest models at the time did not exceed 60% accuracy on questions requiring precise factual judgment.76% similarUnverifiedThe accuracy of the seven models tested on PhysMent ranges from 25% to 67%73% similarUnverifiedAnthropic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%72% similarUnverifiedSelf-Consistency通过多次采样推理路径后投票取多数答案，可进一步提升准确率约5-10个百分点71% similarUnverified研究表明人类评估者在判断LLM输出的事实准确性时，正确率往往只有50%-70%71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/965809API
curl https://kongchang.com/api/v1/knowledge/claims/965809MCP
get_claim(id=965809)