ResearchBenchmarksMultimodal Reasoning on In-domainFollow64Average NMV8.786423.120737.45551.7893Jul 2, 2026Evaluation ResultsMethodMethodLinksAverage NSpeedup FactorAccuracyMVModel=Qwen2.5VL-2.5-7B...Model=Qwen2.5VL-2.5-7B-Instruct, Stopping criteria=N=642026.0764167.3ACModel=Qwen2.5VL-2.5-7B...Model=Qwen2.5VL-2.5-7B-Instruct, Stopping criteria=Adaptive2026.0713.714.6767.4CASModel=Qwen2.5VL-2.5-7B...Model=Qwen2.5VL-2.5-7B-Instruct, Stopping criteria=Confidence-based Adaptive2026.0710.915.8767.5