Pith. sign in

REVIEW

Diffusion Transformer-to-Mamba Distillation for High-Resolution Image Generation

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2506.18999 v1 pith:KLJPPLOD submitted 2025-06-23 cs.CV

Diffusion Transformer-to-Mamba Distillation for High-Resolution Image Generation

classification cs.CV
keywords mambamodeldiffusiongenerationtrainingdistillationhigh-resolutioncomplexity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
abstract

The quadratic computational complexity of self-attention in diffusion transformers (DiT) introduces substantial computational costs in high-resolution image generation. While the linear-complexity Mamba model emerges as a potential alternative, direct Mamba training remains empirically challenging. To address this issue, this paper introduces diffusion transformer-to-mamba distillation (T2MD), forming an efficient training pipeline that facilitates the transition from the self-attention-based transformer to the linear complexity state-space model Mamba. We establish a diffusion self-attention and Mamba hybrid model that simultaneously achieves efficiency and global dependencies. With the proposed layer-level teacher forcing and feature-based knowledge distillation, T2MD alleviates the training difficulty and high cost of a state space model from scratch. Starting from the distilled 512$\times$512 resolution base model, we push the generation towards 2048$\times$2048 images via lightweight adaptation and high-resolution fine-tuning. Experiments demonstrate that our training path leads to low overhead but high-quality text-to-image generation. Importantly, our results also justify the feasibility of using sequential and causal Mamba models for generating non-causal visual output, suggesting the potential for future exploration.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.