Unverified50% confidenceBenchmarkExact time
The study found that the faithfulness score for counterfactual inputs was only 0.05 points lower than for factual inputs (on a 1–5 scale), a difference that is almost negligible.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn scenarios with extreme class imbalance (such as fraud detection), looking solely at accuracy as an evaluation metric is completely meaningless.75% similarUnverifiedThe 2023 TruthfulQA benchmark showed that even the strongest models at the time did not exceed 60% accuracy on questions requiring precise factual judgment.74% similarUnverified在事实性评估任务上模型评判者准确率仍显著低于人类专家，但在风格、连贯性等主观维度已接近人类评审员间一致性水平73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893450API
curl https://kongchang.com/api/v1/knowledge/claims/893450MCP
get_claim(id=893450)