Efficient Agentic Reasoning Through Self-Regulated Simulative Planning

Benchmark Model Rank Results
mathematical-reasoning-on-math-500SR 2 AM-v1.0-30B#1Accuracy: 99.2
multiple-choice-question-answering-mcqa-on-supergpqaSR 2 AM-v1.0-30B#1Accuracy: 73.0
mathematical-reasoning-on-aime24SR 2 AM-v1.0-30B#2Acc: 96.2
question-answering-on-humanity-s-last-exam-hleSR 2 AM-v1.0-30B#11Accuracy: 30.4
mathematical-reasoning-on-aime-2025SR 2 AM-v1.0-30B#18Accuracy: 91.1
deep-research-agents-on-browsecompSR 2 AM-v1.0-30B#60Accuracy: 22.5
multiple-choice-question-answering-mcqa-on-gpqa-diamondSR 2 AM-v1.0-30B#61Accuracy: 77.3