List your company

Logical Reasoning on BIG-Bench Hard Boolean Expression (OOD)

72.4AccuracyCoT-Decoding
61.16864.0846769.916May 7, 2026

Evaluation results

MethodLinks
CoT-DecodingBackbone=mistral-7b2026.0572.4
LOVERBackbone=mistral-7b2026.0572.4
Majority VotingBackbone=mistral-7b2026.0561.6