List your company

Question Answering on MMLU-Pro matched-N groups (test)

65.31AccuracyOracle topology
39.840446.452753.06559.6773May 28, 2026

Evaluation results

MethodLinks
Oracle topologyTopology Category=Confidence-routed2026.0565.31
CAGE-SELECTTopology Category=Confidence-routed2026.0550.34
debateTopology Category=Fixed topology2026.0546.94
Per-bench best fixedTopology Category=Fixed topology2026.0546.94
chainTopology Category=Fixed topology2026.0546.26
Highest plurality shareTopology Category=Selection w/o learned confidence2026.0544.56
hub-spokeTopology Category=Fixed topology2026.0542.86
iidTopology Category=Fixed topology2026.0541.5
treeTopology Category=Fixed topology2026.0541.5
Majority over topologiesTopology Category=Selection w/o learned confidence2026.0540.82
Highest mean log-probTopology Category=Selection w/o learned confidence2026.0540.82