# Intrinsic Distribution Matching sharpens text in two-step image models > A new training method scores 94.25% on Chinese long-text rendering at two network evaluations, with small gains elsewhere and one benchmark it loses. URL: https://claudegpt.si/news/intrinsic-distribution-matching-image-models Published: 2026-10-30T10:52:59.863Z Author: ClaudeGPT Si Newsroom Section: Research Topics: Image generation, Diffusion models, Hugging Face, Research papers Publisher: ClaudeGPT Si ## Key takeaways - IDM drops the separate fake-score network of distribution matching distillation and adds a feature-moment loss. - At two evaluations it scores 94.25% on Chinese and 90.20% on English LongText-Bench, per the authors. - Gains are small in places, and APEX still leads on DPG-Bench in the paper's comparison. Intrinsic Distribution Matching, or IDM, is a training method that makes image generators draw sharper text in two steps instead of dozens. A team led by Deyuan Liu and colleagues describes it in the paper [Intrinsic Distribution Matching](https://arxiv.org/abs/2610.39633), which was the most upvoted paper on Hugging Face's daily list when we checked on October 30, with 231 upvotes. The authors report 94.25% on Chinese and 90.20% on English text rendering at two network evaluations, the highest among the methods they compare. ## The problem: fast image models make more mistakes Most image generators work by removing noise from a picture over many steps. Each step can fix errors from the one before. Few-step models skip most of that, so they are fast but have little room to correct themselves. Long, legible text inside an image is where this hurts. A common way to train fast models is distribution matching distillation, or DMD. It pushes the fast model's output toward the distribution of good reference images. The usual version trains a separate helper network, called a fake-score network, to estimate what the fast model is currently producing. ## What the team changed IDM keeps the full DMD correction. It removes the separate helper network and reads the same signal from two internal branches of the generator itself. It then adds one more training loss that matches the average and the spread of the model's internal features between generated and reference images. The features come from an exponential-moving-average copy of the generator, a slowly updated version of the same model, so no outside image encoder is needed. The authors say their analysis shows these averages and spreads are useful constraints, but that they do not pin down the whole distribution. That is why they keep the DMD correction rather than replacing it. ## The results The team tested on two existing models: Z-Image-Turbo, which has 6 billion parameters, and Qwen-Image, which has 20 billion. - On LongText-Bench at two evaluations, IDM on Z-Image scores 94.25% in Chinese and 90.20% in English. That is 2.63 and 0.32 points above Z-Image-Turbo++, and 3.87 and 1.44 above APEX, per the paper. - On GenEval with Qwen-Image, IDM scores 0.91 against 0.90 for APEX. - On Qwen-Image, FID, a measure of how close images are to real ones, is 6.34 against 6.44 for APEX, at about 0.44 seconds a sample against 0.45. In an ablation on Z-Image, adding the new loss raised GenEval from 77.46 to 77.81. Features from the generator's own copy matched an external DINOv2 encoder, 77.81 against 77.75. ## How much to trust it The gains are often small, sometimes a fraction of a point. APEX scores higher on DPG-Bench, 88.64 against 86.82 in the Z-Image comparison. The ablations ran on one model with full tuning at two evaluations, and the loss weights were fixed rather than tuned. These are the authors' own experiments, and we have not seen independent replication. ## Why it matters If it holds up, IDM is a cheaper recipe for fast image models: fewer moving parts to train and no extra encoder. The code is at [github.com/inclusionAI/IDM](https://github.com/inclusionAI/IDM). The paper's text we read did not clearly say whether trained weights are released. ## Frequently asked questions ### What is Intrinsic Distribution Matching? A training method for one- and few-step image generators. It keeps the standard distillation correction, removes the helper network and adds a loss on feature means and spreads. ### Is the IDM code available? Yes. The paper links code at github.com/inclusionAI/IDM. ### Which models did the authors test? Z-Image-Turbo with 6 billion parameters and Qwen-Image with 20 billion, using full-parameter and LoRA tuning. ## Sources - [Intrinsic Distribution Matching](https://arxiv.org/abs/2610.39633), arXiv - [Intrinsic Distribution Matching (full text)](https://arxiv.org/html/2610.39633), arXiv - [IDM code](https://github.com/inclusionAI/IDM), GitHub (inclusionAI) - [Hugging Face daily papers](https://huggingface.co/papers/2610.39633), Hugging Face