arXiv Machine Learning October 5, 2026 By Wentao Lu, Jesse Clark, Tianyu Zhu

Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs

Read the original on arXiv Machine Learning →

The paper compares two low‑budget methods for converting a 30B Mixture‑of‑Experts autoregressive language model into a diffusion language model. One method updates a subset of the model’s weights in‑place, while the other freezes the context tower and conditions on a frozen causal copy via cross‑attention. With only 1B training tokens, the frozen‑tower approach achieves a HumanEval pass@10 score of 71.60 versus 6.19 for the in‑place method, and retains 95% of the parent’s GSM8K and 99% of its MMLU‑Pro performance.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

arXiv AI

Jun 8

Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation

arXiv:2606. 06712v1 Announce Type: cross Abstract: We study the transformation of autoregressive models (ARLMs) into diffusion language models (DLMs).

By Xingyu Su, Jacob Helwig, Shubham Parashar, Atharv Chagi, Lakshmi Jotsna, Degui Zhi, James Caverlee, Dileep Kalathil, Shuiwang Ji

llms diffusion efficiency

arXiv AI

Jul 17

In-Place Tokenizer Expansion for Pre-trained LLMs

arXiv:2607. 15232v1 Announce Type: cross Abstract: A tokenizer fixed at the start of pre-training allocates vocabulary in proportion to the pre-training corpus, reflecting the deployment priorities at that time.

By Jimmy T. H. Smith, Tarek Dakhran, Alberto Cabrera, Simon S. Lee, Paul Pak, Aditya Tadimeti, Tim Seyde, Maxime Labonne, Alexander Amini, Mathias Lechner

llms rag nlp

arXiv Machine Learning

Jun 11

Teaching Diffusion to Speculate Left-to-Right

arXiv:2606. 11552v1 Announce Type: cross Abstract: Large language models (LLMs) achieve remarkable performance across a wide range of tasks, but their autoregressive decoding process incurs substantial inference costs due to inherently sequential token generation.

By Lexington Whalen, Yuki Ito, Ryo Sakamoto

llms diffusion benchmarks safety

arXiv AI

Jun 2

DLLM-JEPA: Joint Embedding Predictive Architectures for Masked Diffusion Language Models

arXiv:2606. 00091v1 Announce Type: cross Abstract: Joint Embedding Predictive Architectures (JEPAs) have reshaped self-supervised representation learning in vision.

By Sangdae Nam

llms rag diffusion fine-tuning benchmarks

arXiv AI

Jun 29

Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation

arXiv:2606. 27732v1 Announce Type: cross Abstract: Discrete diffusion language models (dLLMs) recover masked tokens in parallel, offering significant speedups over autoregressive (AR) generation.

By Yuhang Chen, Xianfeng Wu, Jinhao Duan, Mingfu Liang, Xiaohan Wei, Yunchen Pu, Fei Tian, Chonglin Sun, Parish Aggarwal, Frank Shyu, Luke Simon, Sandeep Pandey, Xi Liu, Tianlong Chen

llms diffusion efficiency benchmarks

Hugging Face Trending Papers

Jul 23

Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context

Speculative decoding accelerates autoregressive generation by having a cheap draft propose tokens that a target verifies in parallel. Frontier models increasingly ship a built-in Multi-Token-Prediction (MTP/NEXTN) draft head under the assumption that the draft is negligibly cheap.

efficiency

One email a morning, machine-written

Related stories

Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation

In-Place Tokenizer Expansion for Pre-trained LLMs

Teaching Diffusion to Speculate Left-to-Right

DLLM-JEPA: Joint Embedding Predictive Architectures for Masked Diffusion Language Models

Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation

Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context