Interaction between teacher and student model sizes

Characterize how the size of the Qwen3-14B teacher interacts with the size of the student models in MAS-OPD training and performance.

Background

The main experiments hold the teacher fixed at Qwen3-14B while varying the student size between Qwen3-1.7B and Qwen3-4B. This design isolates the effect of student scale but does not determine how teacher scale affects MAS-OPD or how teacher and student capacities interact. The relationship between teacher size and the reported outcomes therefore remains unresolved.

References

how the size of the teacher interacts with what we report is likewise left to future work.

— MAS-OPD: On-Policy Distillation for Multi-agent Systems  (2609.34234 - Zhong et al., 28 Sep 2026) in Appendix, Section 1, “Limitations”