The live demo has been retired — the agent that answered questions is offline.Code, writeup and a recorded demo on GitHub →

Eval scoreboard

Graded against provable FEC ground truth. 24 cases.

Dataset: 470 House districts · 3,039 candidates · 3,949,765 individual contributions · 2024 cycle

100%Accuracy
100%Trajectory
100%Scene
100%Neutrality
0.058Brier
high100% · 16 cases
partial100% · 4 cases
insufficient100% · 4 cases
ItemCorrectTrajectorySceneNeutralConfidence
la01-fundspasspasspasspasshigh
ca20-fundspasspasspasspasshigh
ca17-fundspasspasspasspasshigh
il08-fundspasspasspasspasshigh
mn06-fundspasspasspasspasshigh
wa03-fundspasspasspasspasshigh
co03-fundspasspasspasspasshigh
ca41-fundspasspasspasspasshigh
ny08-fundspasspasspasspasshigh
nj05-fundspasspasspasspasshigh
ny17-fundspasspasspasspasshigh
ak00-fundspasspasspasspasshigh
ca27-fundspasspasspasspasshigh
ny19-fundspasspasspasspasshigh
ny16-fundspasspasspasspasshigh
va07-fundspasspasspasspasshigh
ca53-fundspasspasspasspasspartial
co50-fundspasspasspasspasspartial
ga91-fundspasspasspasspasspartial
ia30-fundspasspasspasspasspartial
az99-nonepasspasspasspassinsufficient
tx99-nonepasspasspasspassinsufficient
ca99-nonepasspasspasspassinsufficient
ny99-nonepasspasspasspassinsufficient