Unverified50% confidenceBenchmarkExact time
In the control group where the answer key was removed from the prompt, the match rate plummeted from 63% to 1% (1/75).
1
Sources
50%
Confidence
Long-term
Relevance
10/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter removing the candidate filtering component, FEVER accuracy dropped to 76.24%.75% similarUnverifiedWhen the incorrect answer key appeared in the prompt, the model's responses matched that incorrect key in 63% (47/75) of cases.70% similarUnverifiedFor agent pairs with large reward gaps, the decision disagreement rate is below 1%.69% similarUnverifiedThe 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.68% similarUnverifiedEvidence-driven pre-improvement validation improved precision from 58.6% to 72.3%, and reduced unsupported objectives per round from 1.93 to 0.50.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/968642API
curl https://kongchang.com/api/v1/knowledge/claims/968642MCP
get_claim(id=968642)