Python 1

Section IV: Dynamic Reasoning Paradigm during Inference 30 Section VII: Reasoning Abilities via Efficient Training Data and Model Compression 16 Section I: RL with Length Reward Design 10 Section II: SFT with Variable-Length CoT Data 8 Section VIII: Evaluation and Benchmark 8 Section III: Compressing Reasoning Steps into Fewer Latent Representation 7 Section VI: Prompts Attribute-Driven Reasoning Routing 5 Section V: Prompt-Guided Efficient Reasoning 4 📢 News 1 Acknowledgment 1

system-2 1 slow-fast 1 self-improve 1 rl 1 reasoning 1 r1 1 prm 1 o3 1 o1 1 mcts 1 macro-action 1 benchmark 1