Add diverse 16-prompt showcase collage to the card

README.md

@@ -36,6 +36,10 @@ pipeline_tag: text-to-image
 | MOSS-Image-SFT (50 steps, CFG 3) | 100 |
 | **This adapter (2 steps, no CFG)** | **2** |
 
+**Showcase — 16 diverse prompts at 2 steps** (comics, calligraphy, photos, posters, anime, game UI; individual images + prompts in [`showcase/`](showcase/)):
+
+![showcase collage](showcase/showcase_grid.jpg)
+
 **Method (IDM, community implementation of [arXiv:2610.39633](https://arxiv.org/abs/2610.39633)):** fake-score network trained with flow matching on generated samples + DMD direction `sg[v_real − v_fake]` + **L_pop EMA feature moment matching** (λ=0.5, mean/std of last-block image-token features vs teacher reference, FIFO queues) — **no adversarial losses, no discriminator**. To our knowledge this is the first open model trained with IDM's loss combination (the official repo is inference-only) and the first purpose-trained 2-step MOSS-Image model. Deviations from the paper: the base is the data-space fixed-point generator ([arXiv:2610.12102](https://arxiv.org/abs/2610.12102)) rather than RCGM; LoRA r128 rather than full-parameter; 249 prompts rather than 5.6M images.
 
 **Training:** 1000 iterations (~2 h on 1×H200), warm-started from darask0's 4-step adapter restricted to 2 step codes, 249 dense-text prompts ([moss-textboost-data](https://huggingface.co/datasets/akhaliq/moss-textboost-data)), EMA 0.998. Shipped checkpoint = **EMA export at iteration 750**, selected by side-by-side eval (below). Trainer: [`training/rapid_moss_train_idm.py`](training/rapid_moss_train_idm.py) (extends darask0's rapid_moss, Apache-2.0). All exports, sample sheets every 100 iterations, and logs: `runs/idm/`.
@@ -179,10 +183,10 @@ image = pipe('A cafe chalkboard menu with the heading "MORNING COFFEE" and three
 
 ## Limitations
 
-- 2-step adapter: trained on 249 text-rendering prompts — expect narrowing toward dense-text/composition prompts; general-purpose quality at 2 steps is good but not exhaustively evaluated. Improvement over darask0's adapter at 2 iterations is incremental. it-1000 exports degraded (L_pop instability late in training) and are kept in `runs/idm/` only for analysis.
+- 2-step adapter: trained on 249 text-rendering prompts — the showcase above suggests good generalization, but general-purpose 2-step quality is not exhaustively evaluated. Improvement over darask0's adapter at 2 iterations is incremental. it-1000 exports degraded (L_pop instability late in training) and are kept in `runs/idm/` only for analysis.
 - v1 experiment: only **100 training images** and 500 steps, and the eval shows no meaningful visible delta over the base at 50 steps.
 - v2/v3 (2-step DiffSynth line): readable composition, illegible text — see v3 above.
-- No LongText-Bench score; evaluation is visual (see `eval/`).
+- No LongText-Bench score; evaluation is visual (see `eval/` and `showcase/`).
 
 ## Citation