AR-MAP: Are Autoregressive Large Language Models Implicit Teachers for Diffusion Large Language Models?

TL;DR

AR-MAP introduces a transfer learning framework where autoregressive language models serve as implicit teachers to improve the preference alignment of diffusion-based large language models, achieving superior performance efficiently.

Contribution

It proposes using autoregressive LLMs as implicit teachers for DLLM alignment, bypassing high variance issues and reducing computational costs.

Findings

Achieves 69.08% average score across tasks

Outperforms existing DLLM alignment methods

Effective knowledge transfer via weight scaling

Abstract

Diffusion Large Language Models (DLLMs) have emerged as a powerful alternative to autoregressive models, enabling parallel token generation across multiple positions. However, preference alignment of DLLMs remains challenging due to high variance introduced by Evidence Lower Bound (ELBO)-based likelihood estimation. In this work, we propose AR-MAP, a novel transfer learning framework that leverages preference-aligned autoregressive LLMs (AR-LLMs) as implicit teachers for DLLM alignment. We reveal that DLLMs can effectively absorb alignment knowledge from AR-LLMs through simple weight scaling, exploiting the shared architectural structure between these divergent generation paradigms. Crucially, our approach circumvents the high variance and computational overhead of direct DLLM alignment and comprehensive experiments across diverse preference alignment tasks demonstrate that AR-MAP achieves competitive or superior performance compared to existing DLLM-specific alignment methods, achieving 69.08\% average score across all tasks and models. Our Code is available at https://github.com/AMAP-ML/AR-MAP.