BREAKING
Arena.ai Adds Factuality Signal
How Factuality Scoring Works
1
Sample battles
↓
2
Extract claims
↓
3
Verify with agents
↓
4
Blend into ranking
0
M+
claims verified
0
Text battles
0
Search battles
0
%
default weight
0
%
Text true-claim rate
0
%
Search true-claim rate
The Leaderboard Reshuffles
Climbs
OpenAI
●
Holds or rises with factuality
●
One model gains ~13 spots
Falls
Meta
●
Drops sharply
●
Some fall over a dozen spots
Popularity vs Correctness
AI NEWS BLITZ
Arena.ai just added a factuality score that's already reshuffling its model leaderboard.