SimpleOPD: Distill long-context reasoning across any tokenizer Using shared text space alignment and KL stabilization, SimpleOPD transfers proof capabilities from SU-01 to Qwen3.5, Intern-S2, GLM,…