Reasoning on BBH matched-N (test)
86.25Accuracy (BBH matched-N)Oracle topology
Evaluation Results
| Method | Links | |
|---|---|---|
| Oracle topologyTopology Category=Confidence-routed2026.05 | 86.25 | |
| CAGE-SELECTTopology Category=Confidence-routed2026.05 | 81.79 | |
| Highest plurality shareTopology Category=Selection w/o learned confidence2026.05 | 78.69 | |
| debateTopology Category=Fixed topology2026.05 | 75.95 | |
| chainTopology Category=Fixed topology2026.05 | 75.95 | |
| Per-bench best fixedTopology Category=Fixed topology2026.05 | 75.95 | |
| Majority over topologiesTopology Category=Selection w/o learned confidence2026.05 | 75.26 | |
| Highest mean log-probTopology Category=Selection w/o learned confidence2026.05 | 74.23 | |
| hub-spokeTopology Category=Fixed topology2026.05 | 72.85 | |
| iidTopology Category=Fixed topology2026.05 | 72.51 | |
| treeTopology Category=Fixed topology2026.05 | 70.1 |