§ qwen · storyline
On-policy distillation in modern LLMs
On-policy distillation emerges as a key post-training technique in models such as Qwen 3.6 and 3.7.
On-policy distillation has emerged as a key post-training technique behind models such as Qwen 3.6 and 3.7.
§ sources1 publication · timeline below