List your company

Multi-hop Commonsense Reasoning on StrategyQA sampled 50 (val)

90AccuracyCoS
84.886.1587.588.85Feb 2, 2026

Evaluation results

MethodLinks
CoSmode=hybrid2026.0290
Directtemperature=02026.0287.5
Chain-of-Thoughttemperature=02026.0287
Structured CoTtemperature=02026.0285
Self-Consistencytemperature=0.7, k=52026.0285