Distribution-Aligned Sequence Distillation for Superior Long-CoT Reasoning

Benchmark Model Rank Results
mathematical-reasoning-on-aime24DASD-4B-Thinking (4B)#15Acc: 88.5
mathematical-reasoning-on-aime-2025DASD-30B-A3B-Thinking-Preview (30B MoE)#27Accuracy: 86.7
code-generation-on-livecodebench-v6DASD-30B-A3B-Thinking-Preview (30B MoE)#28Pass@1: 72.8