# Don't Retrain, Align: Adapting Autoregressive LMs to Diffusion LMs via Representation Alignment

arXiv: https://arxiv.org/abs/2605.06885 · HTML: https://mlsanity.com/paper/2605.06885/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 30 of its 34 references that are in the map.
- First posted on arXiv 7 May 2026.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)

## Builds on (most cited first, 30 of 30)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Deep Unsupervised Learning using Nonequilibrium Thermodynamics](https://mlsanity.com/paper/1503.03585/) (2015)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [XLNet: Generalized Autoregressive Pretraining for Language Understanding](https://mlsanity.com/paper/1906.08237/) (2019)
- [Structured Denoising Diffusion Models in Discrete State-Spaces](https://mlsanity.com/paper/2107.03006/) (2021)
- [Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation](https://mlsanity.com/paper/2305.01210/) (2023)
- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [Diffusion-LM Improves Controllable Text Generation](https://mlsanity.com/paper/2205.14217/) (2022)
- [Simple and Effective Masked Diffusion Language Models](https://mlsanity.com/paper/2406.07524/) (2024)
- [Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution](https://mlsanity.com/paper/2310.16834/) (2023)
- [Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think](https://mlsanity.com/paper/2410.06940/) (2024)
- [Dream 7B: Diffusion Large Language Models](https://mlsanity.com/paper/2508.15487/) (2025)
- [A Continuous Time Framework for Discrete Denoising Models](https://mlsanity.com/paper/2205.14987/) (2022)
- [Scaling Diffusion Language Models via Adaptation from Autoregressive Models](https://mlsanity.com/paper/2410.17891/) (2024)
- [Scaling up Masked Diffusion Models on Text](https://mlsanity.com/paper/2410.18514/) (2024)
- [DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation](https://mlsanity.com/paper/2506.20639/) (2025)
- [Likelihood-Based Diffusion Language Models](https://mlsanity.com/paper/2305.18619/) (2023)
- [NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model](https://mlsanity.com/paper/2508.14444/) (2025)
- [Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think](https://mlsanity.com/paper/2507.01467/) (2025)
- [What matters for Representation Alignment: Global Information or Spatial Structure?](https://mlsanity.com/paper/2512.10794/) (2025)
- [Dream-Coder 7B: An Open Diffusion Language Model for Code](https://mlsanity.com/paper/2509.01142/) (2025)
- [Path Planning for Masked Diffusion Model Sampling](https://mlsanity.com/paper/2502.03540/) (2025)
- [Training and Inference on Any-Order Autoregressive Models the Right Way](https://mlsanity.com/paper/2205.13554/) (2022)
- [No Other Representation Component Is Needed: Diffusion Transformers Can Provide Representation Guidance by Themselves](https://mlsanity.com/paper/2505.02831/) (2025)
- [Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed](https://mlsanity.com/paper/2512.14067/) (2025)
- [Any-Order GPT as Masked Diffusion Model: Decoupling Formulation and Architecture](https://mlsanity.com/paper/2506.19935/) (2025)
- [Planner Aware Path Learning in Diffusion Language Models Training](https://mlsanity.com/paper/2509.23405/) (2025)

## Built on by (most cited first, 6 of 6)

- [Neither Parallel Nor Sequential: How DiffusionGemma Actually Commits Tokens](https://mlsanity.com/paper/2606.14620/) (2026)
- [Distilled Continuous Diffusion Language Models Can Write Code in Few Steps---or One](https://mlsanity.com/paper/2609.04531/) (2026)
- [Continuous Language Diffusion as a Decoder-Interface Problem](https://mlsanity.com/paper/2606.08810/) (2026)
- [ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks](https://mlsanity.com/paper/2609.29102/) (2026)
- [Reasoning with Continuous Latent Diffusion](https://mlsanity.com/paper/2609.35694/) (2026)
- [Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs](https://mlsanity.com/paper/2610.02657/) (2026)

## Abstract (arXiv)

Diffusion language models (DLMs) have recently demonstrated capabilities that complement standard autoregressive (AR) models, particularly in non-sequential generation and bidirectional editing. Although recent work has shown that pretrained autoregressive checkpoints can be converted into diffusion language models, existing recipes primarily transfer parameters through continued denoising training with objective- and attention-level modifications. We instead ask whether the internal representation geometry learned by next-token prediction can be explicitly preserved during AR-to-DLM conversion. We hypothesize that much of the semantic structure learned by AR pretraining can transfer across generation orders, and thus DLM training should be viewed as relearning the decoding path rather than relearning language representations. To investigate this, we introduce REPR-ALIGN, a representation alignment objective that adapts a bidirectional masked diffusion model to reuse representations from a pretrained AR model of identical architecture. Concretely, we align the hidden states of the DLM to the frozen AR model at every layer using cosine similarity, while optimizing the standard masked denoising objective. This simple alignment, with no adapters and no architectural changes beyond the attention mask, yields up to 4x training acceleration in our setting and is particularly effective in low-data regimes. Our results suggest that linguistic representations can transfer across generation order, and that representation alignment provides a simple and effective technique for training diffusion language models. Code is available at https://github.com/pengzhangzhi/Open-dLLM.