Add visual model-card evidence and README banner

README.md

@@ -23,6 +23,8 @@ language:
 
 **Low-bit MOSS-Image for consumer hardware**
 
+<a href="https://huggingface.co/MutantSparrow/LowBitMoss/resolve/main/assets/banner.png"><img src="https://huggingface.co/MutantSparrow/LowBitMoss/resolve/main/assets/banner.png" alt="LowBitMoss example generations across photography, illustration, and text" width="960"></a>
+
 [![GitHub code](https://img.shields.io/badge/GitHub-Code-181717?style=flat-square&logo=github&logoColor=white)](https://github.com/MutantSparrow/LowBitMoss)
 [![Models and data](https://img.shields.io/badge/🤗%20Hugging%20Face-Models%20%26%20Data-ffc107?style=flat-square)](https://huggingface.co/MutantSparrow/LowBitMoss)
 [![Whitepaper](https://img.shields.io/badge/📄%20Paper-Whitepaper-b31b1b?style=flat-square)](https://github.com/MutantSparrow/LowBitMoss/blob/main/paper/whitepaper.pdf)
@@ -36,6 +38,13 @@ The **7.18 GB small distilled package** combines a 2.14 GB DiT, a 4.93 GB pruned
 
 **Author's hands-on validation.** I have also manually benchmarked the models and reviewed their outputs, and I am satisfied with the results. The images have a softer rendering style, but the models retain a surprising amount of visual information and image quality. This is my qualitative assessment from hands-on use; the numerical results in the paper refer to the evaluation configurations described above.
 
+## Visual evidence
+
+Full-resolution visual model cards show example outputs across subjects, styles, text rendering, and complex prompts.
+
+- [mixG_x model card (PNG)](https://huggingface.co/MutantSparrow/LowBitMoss/resolve/main/evidence/model-cards/MODELCARD_mixG_x.png)
+- [mixG_tuned model card (PNG)](https://huggingface.co/MutantSparrow/LowBitMoss/resolve/main/evidence/model-cards/MODELCARD_mixG_tuned.png)
+
 ## Abstract
 
 LowBitMoss explores how mixed-precision quantization can bring MOSS-Image SFT diffusion transformers to consumer hardware. We apply block-wise quantization-aware distillation on a single 16 GB GPU to the 5.08B-parameter DiT, combining learnable group scales, student-stream inputs, and low-rank corrections. Released DiT files occupy 1.72-2.93 GB, compared with 10.16 GB for the bf16 teacher. On 24 prompts at 768 px, the 2.14 GB mixG_tuned model achieves model-judge acceptance of 4.21/5 and letter-exact rendering of all six text prompts. The 1.85 GB and 1.72 GB variants broaden the available storage options.

SHA256SUMS.txt

@@ -3,8 +3,9 @@ a58b01ad36f1644c3ae2b30eee61078e7fccaf442d43b28d58bebb3ba7e66cd1  .gitattributes
 c56b2014c09408266f8059dd25fdcfdd890f468240ce71e7b25a3f1d6210803d  DATA.md
 cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  LICENSE
 3dba711274d65b49a5c2bd92bc0f7cbd23cb7f599d3cb8f590043c92b205f959  NOTICE
-67886f5c44e10b9bcdc410d083be82fbd29d7d01b20c8a5d89284c701741e239  README.md
+805e6e8139c02ddbfaba38135b3b39369a0e04704000c4d5a1ffcfc5cc2df73f  README.md
 71ca2235fd0fecc30822186c59da69ee55b9c369e56e66edf9809d18a8e3fb0b  THIRD_PARTY_NOTICES.md
+579ddf41e80ce1ad2ef764b3a608c19f08afa6a912a491b8ca3b8e7e3a0ecc3d  assets/banner.png
 3b8fbe7d9dc818f1547ef9473720367a71b9663787e0d7488c49911594ba90d4  data/calib_512/p-t0000.pt
 1ce8fc5220e18e87f42124b3ce07ba254e1481c77566f0f25c5b0d5c5f8a8db9  data/calib_512/p-t0001.pt
 23cb854bf7f1c47f01d225b128726c4482e9704c413cd0abe4b790e13e3993ab  data/calib_512/p-t0002.pt
@@ -4019,6 +4020,8 @@ e1de232f64628ef5830c40d20b48307c0cda26dd19b881b0bde14bc3aef720bb  eval/v_ffn_mid
 bc75ad6c7c80975157507b1b9812bf98ec1319125c4f9b6f46bc1e06dd6e2ba4  eval/v_ffn_mid3/latents/10_seed3001.latent
 08bcb025b4bcdfc10b3bc3d5faa183ee0aeee32c16ed91baf40c4969aabe0f71  eval/v_ffn_mid3/latents/14_seed4001.latent
 8c6aa5ef4c5d8462a033867384605dbd0ad9af9b9229039b95d176c2ba5fc636  eval/v_ffn_mid3/latents/15_seed4002.latent
+b191ba5d813fd623303f9903dac5dfc5d98548d2f173dc93a480bd83796dc012  evidence/model-cards/MODELCARD_mixG_tuned.png
+3cfdcac4277a4d766b767f91bb84ee5cea80460ad43e962657dbb2ecf3d99481  evidence/model-cards/MODELCARD_mixG_x.png
 cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  licenses/OpenMOSS-Team--MOSS-Image-SFT--LICENSE
 bbedc3fda3305820b977265f01b8619d87570a6739de3a5582c3464840f1e57a  licenses/Qwen--Qwen3.5-9B--LICENSE
 715266eabd1d6aa98e8f03f6cc90b237ac53ccdcdbb23642e4dfd36a5fc6d6bb  licenses/sources.json
@@ -4065,7 +4068,7 @@ d9d81b071d446cd7391b7c0d156be038f6c6ddf5f273e6b083aa4a9dfd83aa45  paper/whitepap
 c85d3050b830a2a48d1c30ce3cb0bf82b84a2de2ad3bcea753f9fa466cb08298  prompts/calib_prompts.json
 77777cb798f4580f992b8534ca73d4095056183cdba87544b9808d818e50af44  prompts/calib_text_prompts.json
 38debc6129c47f14fe30351d3ed29b1ec1a8b70882489683b0acc66614cb5a1f  prompts/eval_prompts.json
-d0054b9d3bde443d16addee889dd5853c2188761492b54eed481f6a7191c4e9b  release-manifest.json
+bbded9a987c783cb5af2bae6fc8f79d7f9eec96141d197fa1ff44fe89c3758ef  release-manifest.json
 bfd42d73586f52bdb4397e219d9b95441b1af84b3f9212c7286cdc509928d274  text_encoder/qwen3.5_9b_w4a8_gptq_pruned.safetensors
 6689491a55e03345d31e0521965d64ed29187f7768f00d456a9a1181ddd10321  vae/flux2-small-decoder-vae.safetensors
 ea4273f02d1fafbf8e1d1c2cf6018ed8748652eb0bf34f2dd91171f16f15ab62  vae/flux2-vae-small-decoder.safetensors

release-manifest.json

@@ -25,13 +25,17 @@
       "sha256": "3dba711274d65b49a5c2bd92bc0f7cbd23cb7f599d3cb8f590043c92b205f959"
     },
     "README.md": {
-      "bytes": 9119,
-      "sha256": "67886f5c44e10b9bcdc410d083be82fbd29d7d01b20c8a5d89284c701741e239"
+      "bytes": 9817,
+      "sha256": "805e6e8139c02ddbfaba38135b3b39369a0e04704000c4d5a1ffcfc5cc2df73f"
     },
     "THIRD_PARTY_NOTICES.md": {
       "bytes": 1850,
       "sha256": "71ca2235fd0fecc30822186c59da69ee55b9c369e56e66edf9809d18a8e3fb0b"
     },
+    "assets/banner.png": {
+      "bytes": 7617691,
+      "sha256": "579ddf41e80ce1ad2ef764b3a608c19f08afa6a912a491b8ca3b8e7e3a0ecc3d"
+    },
     "data/calib_512/p-t0000.pt": {
       "bytes": 18448879,
       "sha256": "3b8fbe7d9dc818f1547ef9473720367a71b9663787e0d7488c49911594ba90d4"
@@ -16088,6 +16092,14 @@
       "bytes": 1179744,
       "sha256": "8c6aa5ef4c5d8462a033867384605dbd0ad9af9b9229039b95d176c2ba5fc636"
     },
+    "evidence/model-cards/MODELCARD_mixG_tuned.png": {
+      "bytes": 19465245,
+      "sha256": "b191ba5d813fd623303f9903dac5dfc5d98548d2f173dc93a480bd83796dc012"
+    },
+    "evidence/model-cards/MODELCARD_mixG_x.png": {
+      "bytes": 19195520,
+      "sha256": "3cfdcac4277a4d766b767f91bb84ee5cea80460ad43e962657dbb2ecf3d99481"
+    },
     "licenses/OpenMOSS-Team--MOSS-Image-SFT--LICENSE": {
       "bytes": 11358,
       "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30"