ResearchBenchmarksMathematical Reasoning on GSM8K Hard (Det/Cov Metrics)Follow13.9Det. AccuracyRoT-4B8.3889.81911.2512.681Oct 9, 2025Evaluation ResultsMethodMethodLinksDet. AccuracyCov@8Cov@16RoT-4BBackbone=Qwen3-VL, Sam...Backbone=Qwen3-VL, Sampling=MC-dropout2025.1013.914.314.3RoT-2BBackbone=Qwen3-VL, Sam...Backbone=Qwen3-VL, Sampling=MC-dropout2025.108.69.19.2