List your company

Spatial Reasoning on BIG-Bench Hard Navigate (OOD)

57.6AccuracyLOVER
49.2851.4453.655.76May 7, 2026

Evaluation results

MethodLinks
LOVERBackbone=mistral-7b2026.0557.6
CoT-DecodingBackbone=mistral-7b2026.0557.2
Majority VotingBackbone=mistral-7b2026.0549.6