yomi-chen commited on
Commit
15fe5b5
·
verified ·
1 Parent(s): fa43842

Publish run3: six models,1200 outputs and EditableDesignBench v4 evaluation data

Browse files
LICENSE ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ SPDX-License-Identifier: CC-BY-NC-4.0
2
+
3
+ EditableDesignBench
4
+ Copyright (c) 2026 yomi-chen.
5
+
6
+ This dataset is licensed under the Creative Commons Attribution-NonCommercial
7
+ 4.0 International Public License (CC BY-NC 4.0).
8
+
9
+ The full legal terms are available at:
10
+ https://creativecommons.org/licenses/by-nc/4.0/legalcode
11
+
12
+ License summary:
13
+ https://creativecommons.org/licenses/by-nc/4.0/
14
+
15
+ See NOTICE.md for attribution and upstream material notices.
NOTICE.md ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ # Attribution and provenance
2
+
3
+ EditableDesignBench is maintained by yomi-chen and contains newly constructed instruction-guided editing pairs and annotations derived from rendered designs.
4
+
5
+ The source-design prompts are based on BizGenEval, released by Microsoft: https://huggingface.co/datasets/microsoft/BizGenEval. The MIT license notice supplied with the local BizGenEval repository is preserved in `licenses/BizGenEval-MIT.txt`.
6
+
7
+ The task formulation and evaluation convention are inspired by TextEditBench: https://github.com/CSU-JPG/TextEditBench and https://arxiv.org/abs/2512.16270. This is an independently constructed dataset, rather than the official TextEditBench release.
8
+
9
+ EditableDesignBench contributions are distributed under CC BY-NC 4.0. Any independently licensed upstream material retains its applicable license; this release does not replace such rights or imply endorsement by the upstream authors.
README.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: EditableDesignBench Run3 Results
3
+ language:
4
+ - en
5
+ - zh
6
+ license: cc-by-nc-4.0
7
+ task_categories:
8
+ - image-to-image
9
+ tags:
10
+ - image-editing
11
+ - benchmark
12
+ - generated-results
13
+ - EditableDesignBench
14
+ size_categories:
15
+ - 1K<n<10K
16
+ ---
17
+
18
+ # 🎨 EditableDesignBench_result
19
+
20
+ EditableDesignBench **v4.0.0 / run3** 的六个模型结果:每模型 **文字 100 张 + 图片资产 100 张**,共 **1,200 张**。
21
+
22
+ 全部图片与 MSE、PSNR、SSIM、LPIPS 已完成。**尚未运行 MLLM 评分。**
23
+
24
+ [数据集](https://huggingface.co/datasets/yomi-chen/EditableDesignBench) · [评测代码](https://github.com/Ch3nSir/EditableDesignBench)
25
+
26
+ ## 📥 下载并解压
27
+
28
+ ```bash
29
+ pip install -U huggingface_hub
30
+ hf download yomi-chen/EditableDesignBench_result --repo-type dataset --local-dir EditableDesignBench_result
31
+ python EditableDesignBench_result/extract.py
32
+ cd EditableDesignBench_result
33
+ ```
34
+
35
+ 公开下载,无需登录。也可在 Files 页面只下载需要的模型 ZIP;同时下载 `EditableDesignBench_v4_data.zip`,即可取得对应原图、指令和评测标注。
36
+
37
+ | 模型 | 文字 | 资产 | ZIP |
38
+ |---|---:|---:|---|
39
+ | Fibo-2.0-edit-turbo | 100 | 100 | `archives/fibo2-edit-turbo_run3.zip` |
40
+ | JoyAI-Image-Edit-Plus | 100 | 100 | `archives/joyai-image-edit-plus_run3.zip` |
41
+ | Qwen-Image-2.1 | 100 | 100 | `archives/qwen-image21_run3.zip` |
42
+ | Qwen-Image-2.1-Turbo | 100 | 100 | `archives/qwen-image21-turbo_run3.zip` |
43
+ | FLUX 3 Image Klein | 100 | 100 | `archives/flux3-klein_run3.zip` |
44
+ | SenseNova-U1.5-8B-MoT | 100 | 100 | `archives/sensenova-u1.5-8b-mot_run3.zip` |
45
+
46
+ ## 📂 解压后的内容
47
+
48
+ ```text
49
+ data/EditableDesignBench-v4/ # 本次使用的完整 200 条任务:原图、指令、mask、目标图
50
+ text/<model>/predictions/<id>.png # 文字结果的 RGB 评测图片
51
+ assets/<model>/predictions/<id>.png # 资产结果的 RGB 评测图片
52
+ text/<model>/raw_outputs/<id>.png # 模型原始返回文件,逐字节保留
53
+ assets/<model>/raw_outputs/<id>.png
54
+ text/<model>/manifest.jsonl # 逐例指令、路径、尺寸及生成信息
55
+ assets/<model>/manifest.jsonl
56
+ metrics/summary.csv # 两轨道的模型像素指标汇总
57
+ metrics/case_pixel_scores.csv # 1,200 条逐例像素指标
58
+ predictions_index.jsonl # 全部 1,200 条任务与图片的对应关系
59
+ release_info.json # 数据、代码版本和模型配置
60
+ ```
61
+
62
+ `predictions_index.jsonl` 中的路径都相对于本目录,每条记录包含 `source`、`prediction`、`raw_output`、`instruction`、`model`、`track`、`operation` 和尺寸信息。
63
+
64
+ `raw_outputs/` 保留全部原始输出,包括可能存在的 alpha 通道。其余五个模型的 `predictions/` 仅取原始 RGB 通道并去掉已记录的右侧/底部 padding,**不缩放、不合成背景、不重绘**。JoyAI 的 `predictions/` 保留官方分辨率的完整 RGB 输出。
65
+
66
+ 模型 ZIP 中的 `recorded_reports/` 为本次运行的历史生成记录及像素报告,原机器路径已替换成逻辑路径标签,仅供溯源。跨机器评测请使用上述 `predictions/` 与 `manifest.jsonl`,不要直接用历史报告恢复缓存。
67
+
68
+ ## 🧪 使用现有图片做 MLLM 评分
69
+
70
+ 结果包附带本次代码快照 `archives/evaluation_code.zip`,运行 `extract.py` 时会一起解压。无需下载模型权重或重新生成图片。
71
+
72
+ 下面以 Fibo 为例;同样适用于 Qwen-Image-2.1、Qwen-Image-2.1-Turbo、FLUX 3 Klein 和 SenseNova 的原尺寸导出图片:
73
+
74
+ ```bash
75
+ pip install -r evaluation_code/requirements.txt
76
+ export OPENAI_API_KEY="your_api_key"
77
+ export OPENAI_API_URL="https://your-provider.example/v1"
78
+ export OPENAI_MODEL="your-vision-model"
79
+
80
+ # 文字 100 条
81
+ python evaluation_code/scripts/mllm_evaluate.py \
82
+ --dataset data/EditableDesignBench-v4 \
83
+ --predictions text/fibo2-edit-turbo/predictions \
84
+ --output judge/text/fibo2-edit-turbo.json \
85
+ --workers 10 --require-scoring
86
+
87
+ # 资产 100 条
88
+ python evaluation_code/scripts/mllm_evaluate.py \
89
+ --dataset data/EditableDesignBench-v4/assets \
90
+ --predictions assets/fibo2-edit-turbo/predictions \
91
+ --output judge/assets/fibo2-edit-turbo.json \
92
+ --workers 10 --require-scoring
93
+ ```
94
+
95
+ 输入为完整原图、完整结果图和自然语言指令,不提供 mask 或目标图给评委。每例一次请求返回四维评分;重复命令会复用已成功的评分,只补失败/缺失项。本仓库发布时没有进行上述 API 调用。
96
+
97
+ ### JoyAI 的尺寸区别
98
+
99
+ JoyAI 使用官方默认输出档位(1024-base bucket,30 steps,CFG 4),直接输入原图,不指定输出宽高。输出可能与原图尺寸不同;完整变换记录在每条记录的 `official_geometry` 中。
100
+
101
+ **不要将上面的原尺寸评测命令直接用于 JoyAI**:通用脚本要求输入与输出尺寸相同,当前不支持这种 bucket 输出。JoyAI 可用支持不同尺寸图片的评委脚本,以索引里的原图 + 完整结果图 + 原指令评测;请保留两张完整图片,不要为通过尺寸检查而拉伸预测图。其已发布的像素指标使用匹配官方几何变换的参考图与 mask,预测图没有缩放。因此 JoyAI 的像素分数应单独解读。
102
+
103
+ ## 📊 指标与版本
104
+
105
+ - 文字和资产分开统计;四项像素指标均覆盖各模型全部 200 条任务。
106
+ - 像素指标衡量非编辑区域的保持程度,不代表编辑任务已完成。
107
+ - 数据版本:EditableDesignBench `v4.0.0`;生成 seed:每例 `42`。
108
+ - 本次代码版本见 `release_info.json`,评测 prompt 与代码快照一致。
109
+ - 数据集沿用 CC BY-NC 4.0;代码快照中的代码许可证与第三方声明见其 `LICENSE`、`NOTICE.md`。各模型及组件另有各自的上游条款。
archives/EditableDesignBench_v4_data.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:064c51efc0e78f6881782055e7c3a6ca62e92fd4b6b5dfbc62f4feaa7bb014b0
3
+ size 203845340
archives/evaluation_code.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b9a184d57a927e35499d10227a1a31bec5b94f869f91e83c5b5688288ca31a7
3
+ size 249239
archives/fibo2-edit-turbo_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e262f5f648663068acaf3a5e84402694fa696afde3cb6fdc4b9ba159eaa1eca
3
+ size 585272406
archives/flux3-klein_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b5046daf303d63b05434ea8292a32841e6f2fd262b2a9c762e50a496422e75f
3
+ size 492843610
archives/joyai-image-edit-plus_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05d4bd56dfceb63e01c88907cab2ee894c57ed18466d3a4bee48ed674a7d6741
3
+ size 378047485
archives/qwen-image21-turbo_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30865ec4d3120a45996747efdb825e4d30b3d81825ff04c91a65cd52d0a3ebaf
3
+ size 721425031
archives/qwen-image21_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09ae54ee75f7820264b5d00bb1868f1a01d69c094998ae31e6e58203f2266dea
3
+ size 741984060
archives/sensenova-u1.5-8b-mot_run3.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d99b93521693fae4d6ebbb73bcbc46b1c7f0f312180d124a16ca57bee56dc06d
3
+ size 1036278137
extract.py ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Extract downloaded result archives into this directory."""
2
+ from pathlib import Path
3
+ import zipfile
4
+ root=Path(__file__).resolve().parent
5
+ for archive in sorted((root/'archives').glob('*.zip')):
6
+ print('Extracting',archive.name,flush=True)
7
+ with zipfile.ZipFile(archive) as z:
8
+ for member in z.infolist():
9
+ target=(root/member.filename).resolve()
10
+ if not target.is_relative_to(root):raise ValueError('Unsafe archive path')
11
+ z.extractall(root)
12
+ print('Done. See predictions_index.jsonl and README.md.')
files_manifest.json ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": [
3
+ {
4
+ "path": "LICENSE",
5
+ "bytes": 442,
6
+ "sha256": "8d61fb8006d2315e5e0b148c1a26a195c65430377c5f25c67ffbfe434b6e7c99"
7
+ },
8
+ {
9
+ "path": "NOTICE.md",
10
+ "bytes": 921,
11
+ "sha256": "581e56910bf1d6c4ceb3cddb9fbddce50c72faa5ef9c6388264dc8fd6710495d"
12
+ },
13
+ {
14
+ "path": "README.md",
15
+ "bytes": 5727,
16
+ "sha256": "297a5101896ef22978fcd8338af8151de49827f05fbec2375d8301814649c246"
17
+ },
18
+ {
19
+ "path": "archives/EditableDesignBench_v4_data.zip",
20
+ "bytes": 203845340,
21
+ "sha256": "064c51efc0e78f6881782055e7c3a6ca62e92fd4b6b5dfbc62f4feaa7bb014b0"
22
+ },
23
+ {
24
+ "path": "archives/evaluation_code.zip",
25
+ "bytes": 249239,
26
+ "sha256": "1b9a184d57a927e35499d10227a1a31bec5b94f869f91e83c5b5688288ca31a7"
27
+ },
28
+ {
29
+ "path": "archives/fibo2-edit-turbo_run3.zip",
30
+ "bytes": 585272406,
31
+ "sha256": "6e262f5f648663068acaf3a5e84402694fa696afde3cb6fdc4b9ba159eaa1eca"
32
+ },
33
+ {
34
+ "path": "archives/flux3-klein_run3.zip",
35
+ "bytes": 492843610,
36
+ "sha256": "7b5046daf303d63b05434ea8292a32841e6f2fd262b2a9c762e50a496422e75f"
37
+ },
38
+ {
39
+ "path": "archives/joyai-image-edit-plus_run3.zip",
40
+ "bytes": 378047485,
41
+ "sha256": "05d4bd56dfceb63e01c88907cab2ee894c57ed18466d3a4bee48ed674a7d6741"
42
+ },
43
+ {
44
+ "path": "archives/qwen-image21-turbo_run3.zip",
45
+ "bytes": 721425031,
46
+ "sha256": "30865ec4d3120a45996747efdb825e4d30b3d81825ff04c91a65cd52d0a3ebaf"
47
+ },
48
+ {
49
+ "path": "archives/qwen-image21_run3.zip",
50
+ "bytes": 741984060,
51
+ "sha256": "09ae54ee75f7820264b5d00bb1868f1a01d69c094998ae31e6e58203f2266dea"
52
+ },
53
+ {
54
+ "path": "archives/sensenova-u1.5-8b-mot_run3.zip",
55
+ "bytes": 1036278137,
56
+ "sha256": "d99b93521693fae4d6ebbb73bcbc46b1c7f0f312180d124a16ca57bee56dc06d"
57
+ },
58
+ {
59
+ "path": "extract.py",
60
+ "bytes": 543,
61
+ "sha256": "836be7458de0aca0ccc935d1f11892feba2f8c4b00af36459f92ebf51db6defa"
62
+ },
63
+ {
64
+ "path": "metrics/case_pixel_scores.csv",
65
+ "bytes": 133219,
66
+ "sha256": "59a6f537e3483efecd3397dbe0717feae76276fc0d33be57b6381032508a9869"
67
+ },
68
+ {
69
+ "path": "metrics/summary.csv",
70
+ "bytes": 3022,
71
+ "sha256": "124e54c6e5abe798d0e339ddf87729feb711bff737ae9dba3eaf39339002c0d1"
72
+ },
73
+ {
74
+ "path": "metrics/summary.json",
75
+ "bytes": 24054,
76
+ "sha256": "70ff56ea7a9e2867c8ab8a4d176fc5e709ccfd4ee792e49f974e891148322f65"
77
+ },
78
+ {
79
+ "path": "metrics/summary.md",
80
+ "bytes": 2133,
81
+ "sha256": "ff8572e983463aadfbcec6db2355b239fc8cae95b2379e3781aff539e14b856d"
82
+ },
83
+ {
84
+ "path": "predictions_index.jsonl",
85
+ "bytes": 1528524,
86
+ "sha256": "81c6e351f76e3d8084152e8369bc70307e263f50943868a584c73d10e2ba88b5"
87
+ },
88
+ {
89
+ "path": "release_info.json",
90
+ "bytes": 1767,
91
+ "sha256": "fcb5c9e2e86a0f8e754fa80b7249193788b86142197eb9bbc3f1f00062220682"
92
+ }
93
+ ],
94
+ "images": 1200,
95
+ "archives": 8
96
+ }
metrics/case_pixel_scores.csv ADDED
The diff for this file is too large to render. See raw diff
 
metrics/summary.csv ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ track,model,expected,status,image_generation_complete,pixel_complete,full_evaluation_complete,mllm_status,generated,generation_errors,generation_pending,generation_coverage,pixel_valid,pixel_errors,pixel_pending,mse,psnr,ssim,lpips,mse_valid,mse_failed,mse_pending,mse_coverage,psnr_valid,psnr_failed,psnr_pending,psnr_coverage,ssim_valid,ssim_failed,ssim_pending,ssim_coverage,lpips_valid,lpips_failed,lpips_pending,lpips_coverage
2
+ text,fibo2-edit-turbo,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,63.72510300000001,32.614881,0.9887519747018814,0.009714508228935301,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
3
+ text,joyai-image-edit-plus,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,637.828666,23.2513,0.8378133726119995,0.056653072657063604,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
4
+ text,qwen-image21,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,563.82357,21.997068000000002,0.8156222856044769,0.05000474449247122,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
5
+ text,qwen-image21-turbo,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,525.303749,22.352313,0.8266854047775268,0.04349691824987531,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
6
+ text,flux3-klein,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,239.68090900000004,27.698175000000006,0.9418588787317276,0.0269296191399917,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
7
+ text,sensenova-u1.5-8b-mot,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,236.01342200000002,25.889854999999997,0.9430283892154694,0.06245536979287863,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
8
+ assets,fibo2-edit-turbo,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,137.656984,30.427176999999993,0.9824707931280137,0.014146714485250414,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
9
+ assets,joyai-image-edit-plus,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,899.8612939999999,21.741083,0.8341417843103409,0.06435771550051868,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
10
+ assets,qwen-image21,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,534.1776450000001,22.309214,0.8343086391687393,0.04932361725717783,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
11
+ assets,qwen-image21-turbo,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,425.05327400000004,23.320222999999995,0.845193647146225,0.042895335722714666,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
12
+ assets,flux3-klein,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,228.00660699999997,28.060954000000002,0.9503952342271805,0.021453658305108547,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
13
+ assets,sensenova-u1.5-8b-mot,100,pixel_only_complete,True,True,False,not_requested,100,0,0,1.0,100,0,0,308.93107399999997,24.930902999999994,0.9300889307260514,0.06769868858158588,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0,100,0,0,1.0
metrics/summary.json ADDED
@@ -0,0 +1,863 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": 1,
3
+ "updated_at": "2026-11-08T14:28:39.702386+00:00",
4
+ "status": "pixel_only_complete",
5
+ "scoring": "pixel-only",
6
+ "mllm_enabled": false,
7
+ "mllm_status": "not_requested",
8
+ "api_calls_planned": 0,
9
+ "full_evaluation_complete": false,
10
+ "model_configuration_notices": {
11
+ "joyai-image-edit-plus": "Official default bucket output,30steps,CFG4. Original source input without padding; omit width/height. Raw output is preserved. Pixel references/masks are mapped to the official output geometry. These pixel scores use a different resolution protocol from the other five models; report separately."
12
+ },
13
+ "requested_scope_complete": true,
14
+ "image_generation_complete": true,
15
+ "pixel_complete": true,
16
+ "models": [
17
+ "fibo2-edit-turbo",
18
+ "joyai-image-edit-plus",
19
+ "qwen-image21",
20
+ "qwen-image21-turbo",
21
+ "flux3-klein",
22
+ "sensenova-u1.5-8b-mot"
23
+ ],
24
+ "cases_per_model": 200,
25
+ "expected_images": 1200,
26
+ "generated": 1200,
27
+ "generation_errors": 0,
28
+ "generation_pending": 0,
29
+ "pixel_valid": 1200,
30
+ "pixel_errors": 0,
31
+ "pixel_pending": 0,
32
+ "results": [
33
+ {
34
+ "track": "text",
35
+ "model": "fibo2-edit-turbo",
36
+ "expected": 100,
37
+ "status": "pixel_only_complete",
38
+ "scoring": "pixel-only",
39
+ "mllm_status": "not_requested",
40
+ "mllm_executed": false,
41
+ "full_evaluation_complete": false,
42
+ "requested_scope_complete": true,
43
+ "image_generation_complete": true,
44
+ "pixel_complete": true,
45
+ "generated": 100,
46
+ "generation_errors": 0,
47
+ "generation_pending": 0,
48
+ "generation_coverage": 1.0,
49
+ "generation_status": "complete",
50
+ "generation_current": null,
51
+ "pixel_valid": 100,
52
+ "pixel_errors": 0,
53
+ "pixel_pending": 0,
54
+ "pixel_coverage": {
55
+ "mse": {
56
+ "expected": 100,
57
+ "valid": 100,
58
+ "failed": 0,
59
+ "pending": 0,
60
+ "coverage": 1.0,
61
+ "complete": true,
62
+ "validated_by_formal_report": true
63
+ },
64
+ "psnr": {
65
+ "expected": 100,
66
+ "valid": 100,
67
+ "failed": 0,
68
+ "pending": 0,
69
+ "coverage": 1.0,
70
+ "complete": true,
71
+ "validated_by_formal_report": true
72
+ },
73
+ "ssim": {
74
+ "expected": 100,
75
+ "valid": 100,
76
+ "failed": 0,
77
+ "pending": 0,
78
+ "coverage": 1.0,
79
+ "complete": true,
80
+ "validated_by_formal_report": true
81
+ },
82
+ "lpips": {
83
+ "expected": 100,
84
+ "valid": 100,
85
+ "failed": 0,
86
+ "pending": 0,
87
+ "coverage": 1.0,
88
+ "complete": true,
89
+ "validated_by_formal_report": true
90
+ }
91
+ },
92
+ "metrics": {
93
+ "mse": 63.72510300000001,
94
+ "psnr": 32.614881,
95
+ "ssim": 0.9887519747018814,
96
+ "lpips": 0.009714508228935301
97
+ },
98
+ "progress_error": null,
99
+ "validation_errors": [],
100
+ "run_dir": "experiment/output/run3/text/fibo2-edit-turbo"
101
+ },
102
+ {
103
+ "track": "text",
104
+ "model": "joyai-image-edit-plus",
105
+ "expected": 100,
106
+ "status": "pixel_only_complete",
107
+ "scoring": "pixel-only",
108
+ "mllm_status": "not_requested",
109
+ "mllm_executed": false,
110
+ "full_evaluation_complete": false,
111
+ "requested_scope_complete": true,
112
+ "image_generation_complete": true,
113
+ "pixel_complete": true,
114
+ "generated": 100,
115
+ "generation_errors": 0,
116
+ "generation_pending": 0,
117
+ "generation_coverage": 1.0,
118
+ "generation_status": "complete",
119
+ "generation_current": null,
120
+ "pixel_valid": 100,
121
+ "pixel_errors": 0,
122
+ "pixel_pending": 0,
123
+ "pixel_coverage": {
124
+ "mse": {
125
+ "expected": 100,
126
+ "valid": 100,
127
+ "failed": 0,
128
+ "pending": 0,
129
+ "coverage": 1.0,
130
+ "complete": true,
131
+ "validated_by_formal_report": true
132
+ },
133
+ "psnr": {
134
+ "expected": 100,
135
+ "valid": 100,
136
+ "failed": 0,
137
+ "pending": 0,
138
+ "coverage": 1.0,
139
+ "complete": true,
140
+ "validated_by_formal_report": true
141
+ },
142
+ "ssim": {
143
+ "expected": 100,
144
+ "valid": 100,
145
+ "failed": 0,
146
+ "pending": 0,
147
+ "coverage": 1.0,
148
+ "complete": true,
149
+ "validated_by_formal_report": true
150
+ },
151
+ "lpips": {
152
+ "expected": 100,
153
+ "valid": 100,
154
+ "failed": 0,
155
+ "pending": 0,
156
+ "coverage": 1.0,
157
+ "complete": true,
158
+ "validated_by_formal_report": true
159
+ }
160
+ },
161
+ "metrics": {
162
+ "mse": 637.828666,
163
+ "psnr": 23.2513,
164
+ "ssim": 0.8378133726119995,
165
+ "lpips": 0.056653072657063604
166
+ },
167
+ "progress_error": null,
168
+ "validation_errors": [],
169
+ "run_dir": "experiment/output/run3/text/joyai-image-edit-plus"
170
+ },
171
+ {
172
+ "track": "text",
173
+ "model": "qwen-image21",
174
+ "expected": 100,
175
+ "status": "pixel_only_complete",
176
+ "scoring": "pixel-only",
177
+ "mllm_status": "not_requested",
178
+ "mllm_executed": false,
179
+ "full_evaluation_complete": false,
180
+ "requested_scope_complete": true,
181
+ "image_generation_complete": true,
182
+ "pixel_complete": true,
183
+ "generated": 100,
184
+ "generation_errors": 0,
185
+ "generation_pending": 0,
186
+ "generation_coverage": 1.0,
187
+ "generation_status": "complete",
188
+ "generation_current": null,
189
+ "pixel_valid": 100,
190
+ "pixel_errors": 0,
191
+ "pixel_pending": 0,
192
+ "pixel_coverage": {
193
+ "mse": {
194
+ "expected": 100,
195
+ "valid": 100,
196
+ "failed": 0,
197
+ "pending": 0,
198
+ "coverage": 1.0,
199
+ "complete": true,
200
+ "validated_by_formal_report": true
201
+ },
202
+ "psnr": {
203
+ "expected": 100,
204
+ "valid": 100,
205
+ "failed": 0,
206
+ "pending": 0,
207
+ "coverage": 1.0,
208
+ "complete": true,
209
+ "validated_by_formal_report": true
210
+ },
211
+ "ssim": {
212
+ "expected": 100,
213
+ "valid": 100,
214
+ "failed": 0,
215
+ "pending": 0,
216
+ "coverage": 1.0,
217
+ "complete": true,
218
+ "validated_by_formal_report": true
219
+ },
220
+ "lpips": {
221
+ "expected": 100,
222
+ "valid": 100,
223
+ "failed": 0,
224
+ "pending": 0,
225
+ "coverage": 1.0,
226
+ "complete": true,
227
+ "validated_by_formal_report": true
228
+ }
229
+ },
230
+ "metrics": {
231
+ "mse": 563.82357,
232
+ "psnr": 21.997068000000002,
233
+ "ssim": 0.8156222856044769,
234
+ "lpips": 0.05000474449247122
235
+ },
236
+ "progress_error": null,
237
+ "validation_errors": [],
238
+ "run_dir": "experiment/output/run3/text/qwen-image21"
239
+ },
240
+ {
241
+ "track": "text",
242
+ "model": "qwen-image21-turbo",
243
+ "expected": 100,
244
+ "status": "pixel_only_complete",
245
+ "scoring": "pixel-only",
246
+ "mllm_status": "not_requested",
247
+ "mllm_executed": false,
248
+ "full_evaluation_complete": false,
249
+ "requested_scope_complete": true,
250
+ "image_generation_complete": true,
251
+ "pixel_complete": true,
252
+ "generated": 100,
253
+ "generation_errors": 0,
254
+ "generation_pending": 0,
255
+ "generation_coverage": 1.0,
256
+ "generation_status": "complete",
257
+ "generation_current": null,
258
+ "pixel_valid": 100,
259
+ "pixel_errors": 0,
260
+ "pixel_pending": 0,
261
+ "pixel_coverage": {
262
+ "mse": {
263
+ "expected": 100,
264
+ "valid": 100,
265
+ "failed": 0,
266
+ "pending": 0,
267
+ "coverage": 1.0,
268
+ "complete": true,
269
+ "validated_by_formal_report": true
270
+ },
271
+ "psnr": {
272
+ "expected": 100,
273
+ "valid": 100,
274
+ "failed": 0,
275
+ "pending": 0,
276
+ "coverage": 1.0,
277
+ "complete": true,
278
+ "validated_by_formal_report": true
279
+ },
280
+ "ssim": {
281
+ "expected": 100,
282
+ "valid": 100,
283
+ "failed": 0,
284
+ "pending": 0,
285
+ "coverage": 1.0,
286
+ "complete": true,
287
+ "validated_by_formal_report": true
288
+ },
289
+ "lpips": {
290
+ "expected": 100,
291
+ "valid": 100,
292
+ "failed": 0,
293
+ "pending": 0,
294
+ "coverage": 1.0,
295
+ "complete": true,
296
+ "validated_by_formal_report": true
297
+ }
298
+ },
299
+ "metrics": {
300
+ "mse": 525.303749,
301
+ "psnr": 22.352313,
302
+ "ssim": 0.8266854047775268,
303
+ "lpips": 0.04349691824987531
304
+ },
305
+ "progress_error": null,
306
+ "validation_errors": [],
307
+ "run_dir": "experiment/output/run3/text/qwen-image21-turbo"
308
+ },
309
+ {
310
+ "track": "text",
311
+ "model": "flux3-klein",
312
+ "expected": 100,
313
+ "status": "pixel_only_complete",
314
+ "scoring": "pixel-only",
315
+ "mllm_status": "not_requested",
316
+ "mllm_executed": false,
317
+ "full_evaluation_complete": false,
318
+ "requested_scope_complete": true,
319
+ "image_generation_complete": true,
320
+ "pixel_complete": true,
321
+ "generated": 100,
322
+ "generation_errors": 0,
323
+ "generation_pending": 0,
324
+ "generation_coverage": 1.0,
325
+ "generation_status": "complete",
326
+ "generation_current": null,
327
+ "pixel_valid": 100,
328
+ "pixel_errors": 0,
329
+ "pixel_pending": 0,
330
+ "pixel_coverage": {
331
+ "mse": {
332
+ "expected": 100,
333
+ "valid": 100,
334
+ "failed": 0,
335
+ "pending": 0,
336
+ "coverage": 1.0,
337
+ "complete": true,
338
+ "validated_by_formal_report": true
339
+ },
340
+ "psnr": {
341
+ "expected": 100,
342
+ "valid": 100,
343
+ "failed": 0,
344
+ "pending": 0,
345
+ "coverage": 1.0,
346
+ "complete": true,
347
+ "validated_by_formal_report": true
348
+ },
349
+ "ssim": {
350
+ "expected": 100,
351
+ "valid": 100,
352
+ "failed": 0,
353
+ "pending": 0,
354
+ "coverage": 1.0,
355
+ "complete": true,
356
+ "validated_by_formal_report": true
357
+ },
358
+ "lpips": {
359
+ "expected": 100,
360
+ "valid": 100,
361
+ "failed": 0,
362
+ "pending": 0,
363
+ "coverage": 1.0,
364
+ "complete": true,
365
+ "validated_by_formal_report": true
366
+ }
367
+ },
368
+ "metrics": {
369
+ "mse": 239.68090900000004,
370
+ "psnr": 27.698175000000006,
371
+ "ssim": 0.9418588787317276,
372
+ "lpips": 0.0269296191399917
373
+ },
374
+ "progress_error": null,
375
+ "validation_errors": [],
376
+ "run_dir": "experiment/output/run3/text/flux3-klein"
377
+ },
378
+ {
379
+ "track": "text",
380
+ "model": "sensenova-u1.5-8b-mot",
381
+ "expected": 100,
382
+ "status": "pixel_only_complete",
383
+ "scoring": "pixel-only",
384
+ "mllm_status": "not_requested",
385
+ "mllm_executed": false,
386
+ "full_evaluation_complete": false,
387
+ "requested_scope_complete": true,
388
+ "image_generation_complete": true,
389
+ "pixel_complete": true,
390
+ "generated": 100,
391
+ "generation_errors": 0,
392
+ "generation_pending": 0,
393
+ "generation_coverage": 1.0,
394
+ "generation_status": "complete",
395
+ "generation_current": null,
396
+ "pixel_valid": 100,
397
+ "pixel_errors": 0,
398
+ "pixel_pending": 0,
399
+ "pixel_coverage": {
400
+ "mse": {
401
+ "expected": 100,
402
+ "valid": 100,
403
+ "failed": 0,
404
+ "pending": 0,
405
+ "coverage": 1.0,
406
+ "complete": true,
407
+ "validated_by_formal_report": true
408
+ },
409
+ "psnr": {
410
+ "expected": 100,
411
+ "valid": 100,
412
+ "failed": 0,
413
+ "pending": 0,
414
+ "coverage": 1.0,
415
+ "complete": true,
416
+ "validated_by_formal_report": true
417
+ },
418
+ "ssim": {
419
+ "expected": 100,
420
+ "valid": 100,
421
+ "failed": 0,
422
+ "pending": 0,
423
+ "coverage": 1.0,
424
+ "complete": true,
425
+ "validated_by_formal_report": true
426
+ },
427
+ "lpips": {
428
+ "expected": 100,
429
+ "valid": 100,
430
+ "failed": 0,
431
+ "pending": 0,
432
+ "coverage": 1.0,
433
+ "complete": true,
434
+ "validated_by_formal_report": true
435
+ }
436
+ },
437
+ "metrics": {
438
+ "mse": 236.01342200000002,
439
+ "psnr": 25.889854999999997,
440
+ "ssim": 0.9430283892154694,
441
+ "lpips": 0.06245536979287863
442
+ },
443
+ "progress_error": null,
444
+ "validation_errors": [],
445
+ "run_dir": "experiment/output/run3/text/sensenova-u1.5-8b-mot"
446
+ },
447
+ {
448
+ "track": "assets",
449
+ "model": "fibo2-edit-turbo",
450
+ "expected": 100,
451
+ "status": "pixel_only_complete",
452
+ "scoring": "pixel-only",
453
+ "mllm_status": "not_requested",
454
+ "mllm_executed": false,
455
+ "full_evaluation_complete": false,
456
+ "requested_scope_complete": true,
457
+ "image_generation_complete": true,
458
+ "pixel_complete": true,
459
+ "generated": 100,
460
+ "generation_errors": 0,
461
+ "generation_pending": 0,
462
+ "generation_coverage": 1.0,
463
+ "generation_status": "complete",
464
+ "generation_current": null,
465
+ "pixel_valid": 100,
466
+ "pixel_errors": 0,
467
+ "pixel_pending": 0,
468
+ "pixel_coverage": {
469
+ "mse": {
470
+ "expected": 100,
471
+ "valid": 100,
472
+ "failed": 0,
473
+ "pending": 0,
474
+ "coverage": 1.0,
475
+ "complete": true,
476
+ "validated_by_formal_report": true
477
+ },
478
+ "psnr": {
479
+ "expected": 100,
480
+ "valid": 100,
481
+ "failed": 0,
482
+ "pending": 0,
483
+ "coverage": 1.0,
484
+ "complete": true,
485
+ "validated_by_formal_report": true
486
+ },
487
+ "ssim": {
488
+ "expected": 100,
489
+ "valid": 100,
490
+ "failed": 0,
491
+ "pending": 0,
492
+ "coverage": 1.0,
493
+ "complete": true,
494
+ "validated_by_formal_report": true
495
+ },
496
+ "lpips": {
497
+ "expected": 100,
498
+ "valid": 100,
499
+ "failed": 0,
500
+ "pending": 0,
501
+ "coverage": 1.0,
502
+ "complete": true,
503
+ "validated_by_formal_report": true
504
+ }
505
+ },
506
+ "metrics": {
507
+ "mse": 137.656984,
508
+ "psnr": 30.427176999999993,
509
+ "ssim": 0.9824707931280137,
510
+ "lpips": 0.014146714485250414
511
+ },
512
+ "progress_error": null,
513
+ "validation_errors": [],
514
+ "run_dir": "experiment/output/run3/assets/fibo2-edit-turbo"
515
+ },
516
+ {
517
+ "track": "assets",
518
+ "model": "joyai-image-edit-plus",
519
+ "expected": 100,
520
+ "status": "pixel_only_complete",
521
+ "scoring": "pixel-only",
522
+ "mllm_status": "not_requested",
523
+ "mllm_executed": false,
524
+ "full_evaluation_complete": false,
525
+ "requested_scope_complete": true,
526
+ "image_generation_complete": true,
527
+ "pixel_complete": true,
528
+ "generated": 100,
529
+ "generation_errors": 0,
530
+ "generation_pending": 0,
531
+ "generation_coverage": 1.0,
532
+ "generation_status": "complete",
533
+ "generation_current": null,
534
+ "pixel_valid": 100,
535
+ "pixel_errors": 0,
536
+ "pixel_pending": 0,
537
+ "pixel_coverage": {
538
+ "mse": {
539
+ "expected": 100,
540
+ "valid": 100,
541
+ "failed": 0,
542
+ "pending": 0,
543
+ "coverage": 1.0,
544
+ "complete": true,
545
+ "validated_by_formal_report": true
546
+ },
547
+ "psnr": {
548
+ "expected": 100,
549
+ "valid": 100,
550
+ "failed": 0,
551
+ "pending": 0,
552
+ "coverage": 1.0,
553
+ "complete": true,
554
+ "validated_by_formal_report": true
555
+ },
556
+ "ssim": {
557
+ "expected": 100,
558
+ "valid": 100,
559
+ "failed": 0,
560
+ "pending": 0,
561
+ "coverage": 1.0,
562
+ "complete": true,
563
+ "validated_by_formal_report": true
564
+ },
565
+ "lpips": {
566
+ "expected": 100,
567
+ "valid": 100,
568
+ "failed": 0,
569
+ "pending": 0,
570
+ "coverage": 1.0,
571
+ "complete": true,
572
+ "validated_by_formal_report": true
573
+ }
574
+ },
575
+ "metrics": {
576
+ "mse": 899.8612939999999,
577
+ "psnr": 21.741083,
578
+ "ssim": 0.8341417843103409,
579
+ "lpips": 0.06435771550051868
580
+ },
581
+ "progress_error": null,
582
+ "validation_errors": [],
583
+ "run_dir": "experiment/output/run3/assets/joyai-image-edit-plus"
584
+ },
585
+ {
586
+ "track": "assets",
587
+ "model": "qwen-image21",
588
+ "expected": 100,
589
+ "status": "pixel_only_complete",
590
+ "scoring": "pixel-only",
591
+ "mllm_status": "not_requested",
592
+ "mllm_executed": false,
593
+ "full_evaluation_complete": false,
594
+ "requested_scope_complete": true,
595
+ "image_generation_complete": true,
596
+ "pixel_complete": true,
597
+ "generated": 100,
598
+ "generation_errors": 0,
599
+ "generation_pending": 0,
600
+ "generation_coverage": 1.0,
601
+ "generation_status": "complete",
602
+ "generation_current": null,
603
+ "pixel_valid": 100,
604
+ "pixel_errors": 0,
605
+ "pixel_pending": 0,
606
+ "pixel_coverage": {
607
+ "mse": {
608
+ "expected": 100,
609
+ "valid": 100,
610
+ "failed": 0,
611
+ "pending": 0,
612
+ "coverage": 1.0,
613
+ "complete": true,
614
+ "validated_by_formal_report": true
615
+ },
616
+ "psnr": {
617
+ "expected": 100,
618
+ "valid": 100,
619
+ "failed": 0,
620
+ "pending": 0,
621
+ "coverage": 1.0,
622
+ "complete": true,
623
+ "validated_by_formal_report": true
624
+ },
625
+ "ssim": {
626
+ "expected": 100,
627
+ "valid": 100,
628
+ "failed": 0,
629
+ "pending": 0,
630
+ "coverage": 1.0,
631
+ "complete": true,
632
+ "validated_by_formal_report": true
633
+ },
634
+ "lpips": {
635
+ "expected": 100,
636
+ "valid": 100,
637
+ "failed": 0,
638
+ "pending": 0,
639
+ "coverage": 1.0,
640
+ "complete": true,
641
+ "validated_by_formal_report": true
642
+ }
643
+ },
644
+ "metrics": {
645
+ "mse": 534.1776450000001,
646
+ "psnr": 22.309214,
647
+ "ssim": 0.8343086391687393,
648
+ "lpips": 0.04932361725717783
649
+ },
650
+ "progress_error": null,
651
+ "validation_errors": [],
652
+ "run_dir": "experiment/output/run3/assets/qwen-image21"
653
+ },
654
+ {
655
+ "track": "assets",
656
+ "model": "qwen-image21-turbo",
657
+ "expected": 100,
658
+ "status": "pixel_only_complete",
659
+ "scoring": "pixel-only",
660
+ "mllm_status": "not_requested",
661
+ "mllm_executed": false,
662
+ "full_evaluation_complete": false,
663
+ "requested_scope_complete": true,
664
+ "image_generation_complete": true,
665
+ "pixel_complete": true,
666
+ "generated": 100,
667
+ "generation_errors": 0,
668
+ "generation_pending": 0,
669
+ "generation_coverage": 1.0,
670
+ "generation_status": "complete",
671
+ "generation_current": null,
672
+ "pixel_valid": 100,
673
+ "pixel_errors": 0,
674
+ "pixel_pending": 0,
675
+ "pixel_coverage": {
676
+ "mse": {
677
+ "expected": 100,
678
+ "valid": 100,
679
+ "failed": 0,
680
+ "pending": 0,
681
+ "coverage": 1.0,
682
+ "complete": true,
683
+ "validated_by_formal_report": true
684
+ },
685
+ "psnr": {
686
+ "expected": 100,
687
+ "valid": 100,
688
+ "failed": 0,
689
+ "pending": 0,
690
+ "coverage": 1.0,
691
+ "complete": true,
692
+ "validated_by_formal_report": true
693
+ },
694
+ "ssim": {
695
+ "expected": 100,
696
+ "valid": 100,
697
+ "failed": 0,
698
+ "pending": 0,
699
+ "coverage": 1.0,
700
+ "complete": true,
701
+ "validated_by_formal_report": true
702
+ },
703
+ "lpips": {
704
+ "expected": 100,
705
+ "valid": 100,
706
+ "failed": 0,
707
+ "pending": 0,
708
+ "coverage": 1.0,
709
+ "complete": true,
710
+ "validated_by_formal_report": true
711
+ }
712
+ },
713
+ "metrics": {
714
+ "mse": 425.05327400000004,
715
+ "psnr": 23.320222999999995,
716
+ "ssim": 0.845193647146225,
717
+ "lpips": 0.042895335722714666
718
+ },
719
+ "progress_error": null,
720
+ "validation_errors": [],
721
+ "run_dir": "experiment/output/run3/assets/qwen-image21-turbo"
722
+ },
723
+ {
724
+ "track": "assets",
725
+ "model": "flux3-klein",
726
+ "expected": 100,
727
+ "status": "pixel_only_complete",
728
+ "scoring": "pixel-only",
729
+ "mllm_status": "not_requested",
730
+ "mllm_executed": false,
731
+ "full_evaluation_complete": false,
732
+ "requested_scope_complete": true,
733
+ "image_generation_complete": true,
734
+ "pixel_complete": true,
735
+ "generated": 100,
736
+ "generation_errors": 0,
737
+ "generation_pending": 0,
738
+ "generation_coverage": 1.0,
739
+ "generation_status": "complete",
740
+ "generation_current": null,
741
+ "pixel_valid": 100,
742
+ "pixel_errors": 0,
743
+ "pixel_pending": 0,
744
+ "pixel_coverage": {
745
+ "mse": {
746
+ "expected": 100,
747
+ "valid": 100,
748
+ "failed": 0,
749
+ "pending": 0,
750
+ "coverage": 1.0,
751
+ "complete": true,
752
+ "validated_by_formal_report": true
753
+ },
754
+ "psnr": {
755
+ "expected": 100,
756
+ "valid": 100,
757
+ "failed": 0,
758
+ "pending": 0,
759
+ "coverage": 1.0,
760
+ "complete": true,
761
+ "validated_by_formal_report": true
762
+ },
763
+ "ssim": {
764
+ "expected": 100,
765
+ "valid": 100,
766
+ "failed": 0,
767
+ "pending": 0,
768
+ "coverage": 1.0,
769
+ "complete": true,
770
+ "validated_by_formal_report": true
771
+ },
772
+ "lpips": {
773
+ "expected": 100,
774
+ "valid": 100,
775
+ "failed": 0,
776
+ "pending": 0,
777
+ "coverage": 1.0,
778
+ "complete": true,
779
+ "validated_by_formal_report": true
780
+ }
781
+ },
782
+ "metrics": {
783
+ "mse": 228.00660699999997,
784
+ "psnr": 28.060954000000002,
785
+ "ssim": 0.9503952342271805,
786
+ "lpips": 0.021453658305108547
787
+ },
788
+ "progress_error": null,
789
+ "validation_errors": [],
790
+ "run_dir": "experiment/output/run3/assets/flux3-klein"
791
+ },
792
+ {
793
+ "track": "assets",
794
+ "model": "sensenova-u1.5-8b-mot",
795
+ "expected": 100,
796
+ "status": "pixel_only_complete",
797
+ "scoring": "pixel-only",
798
+ "mllm_status": "not_requested",
799
+ "mllm_executed": false,
800
+ "full_evaluation_complete": false,
801
+ "requested_scope_complete": true,
802
+ "image_generation_complete": true,
803
+ "pixel_complete": true,
804
+ "generated": 100,
805
+ "generation_errors": 0,
806
+ "generation_pending": 0,
807
+ "generation_coverage": 1.0,
808
+ "generation_status": "complete",
809
+ "generation_current": null,
810
+ "pixel_valid": 100,
811
+ "pixel_errors": 0,
812
+ "pixel_pending": 0,
813
+ "pixel_coverage": {
814
+ "mse": {
815
+ "expected": 100,
816
+ "valid": 100,
817
+ "failed": 0,
818
+ "pending": 0,
819
+ "coverage": 1.0,
820
+ "complete": true,
821
+ "validated_by_formal_report": true
822
+ },
823
+ "psnr": {
824
+ "expected": 100,
825
+ "valid": 100,
826
+ "failed": 0,
827
+ "pending": 0,
828
+ "coverage": 1.0,
829
+ "complete": true,
830
+ "validated_by_formal_report": true
831
+ },
832
+ "ssim": {
833
+ "expected": 100,
834
+ "valid": 100,
835
+ "failed": 0,
836
+ "pending": 0,
837
+ "coverage": 1.0,
838
+ "complete": true,
839
+ "validated_by_formal_report": true
840
+ },
841
+ "lpips": {
842
+ "expected": 100,
843
+ "valid": 100,
844
+ "failed": 0,
845
+ "pending": 0,
846
+ "coverage": 1.0,
847
+ "complete": true,
848
+ "validated_by_formal_report": true
849
+ }
850
+ },
851
+ "metrics": {
852
+ "mse": 308.93107399999997,
853
+ "psnr": 24.930902999999994,
854
+ "ssim": 0.9300889307260514,
855
+ "lpips": 0.06769868858158588
856
+ },
857
+ "progress_error": null,
858
+ "validation_errors": [],
859
+ "run_dir": "experiment/output/run3/assets/sensenova-u1.5-8b-mot"
860
+ }
861
+ ],
862
+ "note": "This run generates images and evaluates pixel preservation only. MLLM judging was not requested or executed; semantic scores and joint qualification are unavailable. Completion refers only to all planned images and pixel metrics, never to full semantic benchmark evaluation."
863
+ }
metrics/summary.md ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Run3: generation and pixel metrics
2
+
3
+ **Pixel-only. MLLM judging is not requested or executed. No semantic scores or joint qualification are reported.**
4
+
5
+ Generated: 1200/1200; generation errors: 0; pending: 0.
6
+ Pixel-valid cases: 1200/1200. Requested image/pixel scope complete: True.
7
+
8
+ Pixel means appear only after the formal report validates full coverage for that metric. Text and assets are reported separately.
9
+
10
+ **Configuration notice — joyai-image-edit-plus:** Official default bucket output,30steps,CFG4. Original source input without padding; omit width/height. Raw output is preserved. Pixel references/masks are mapped to the official output geometry. These pixel scores use a different resolution protocol from the other five models; report separately.
11
+
12
+ ## text
13
+
14
+ | Model | Images | Errors | Pending | Pixel valid | MSE ↓ | PSNR ↑ | SSIM ↑ | LPIPS ↓ |
15
+ |---|---:|---:|---:|---:|---:|---:|---:|---:|
16
+ | fibo2-edit-turbo | 100/100 | 0 | 0 | 100/100 | 63.7251 | 32.6149 | 0.9888 | 0.0097 |
17
+ | joyai-image-edit-plus | 100/100 | 0 | 0 | 100/100 | 637.8287 | 23.2513 | 0.8378 | 0.0567 |
18
+ | qwen-image21 | 100/100 | 0 | 0 | 100/100 | 563.8236 | 21.9971 | 0.8156 | 0.0500 |
19
+ | qwen-image21-turbo | 100/100 | 0 | 0 | 100/100 | 525.3037 | 22.3523 | 0.8267 | 0.0435 |
20
+ | flux3-klein | 100/100 | 0 | 0 | 100/100 | 239.6809 | 27.6982 | 0.9419 | 0.0269 |
21
+ | sensenova-u1.5-8b-mot | 100/100 | 0 | 0 | 100/100 | 236.0134 | 25.8899 | 0.9430 | 0.0625 |
22
+
23
+ ## assets
24
+
25
+ | Model | Images | Errors | Pending | Pixel valid | MSE ↓ | PSNR ↑ | SSIM ↑ | LPIPS ↓ |
26
+ |---|---:|---:|---:|---:|---:|---:|---:|---:|
27
+ | fibo2-edit-turbo | 100/100 | 0 | 0 | 100/100 | 137.6570 | 30.4272 | 0.9825 | 0.0141 |
28
+ | joyai-image-edit-plus | 100/100 | 0 | 0 | 100/100 | 899.8613 | 21.7411 | 0.8341 | 0.0644 |
29
+ | qwen-image21 | 100/100 | 0 | 0 | 100/100 | 534.1776 | 22.3092 | 0.8343 | 0.0493 |
30
+ | qwen-image21-turbo | 100/100 | 0 | 0 | 100/100 | 425.0533 | 23.3202 | 0.8452 | 0.0429 |
31
+ | flux3-klein | 100/100 | 0 | 0 | 100/100 | 228.0066 | 28.0610 | 0.9504 | 0.0215 |
32
+ | sensenova-u1.5-8b-mot | 100/100 | 0 | 0 | 100/100 | 308.9311 | 24.9309 | 0.9301 | 0.0677 |
33
+
predictions_index.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
release_info.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run": "run3",
3
+ "dataset": "yomi-chen/EditableDesignBench",
4
+ "dataset_revision": "v4.0.0",
5
+ "dataset_commit": "6b2b9258f50967d6f179325c5bca99ed88ce6e32",
6
+ "code_repository": "https://github.com/Ch3nSir/EditableDesignBench",
7
+ "code_commit": "8e087fccdc22798b9b131b62136314c943a2fecf",
8
+ "models": [
9
+ "fibo2-edit-turbo",
10
+ "joyai-image-edit-plus",
11
+ "qwen-image21",
12
+ "qwen-image21-turbo",
13
+ "flux3-klein",
14
+ "sensenova-u1.5-8b-mot"
15
+ ],
16
+ "text_cases": 100,
17
+ "asset_cases": 100,
18
+ "images": 1200,
19
+ "mllm_scored": false,
20
+ "pixel_metrics": [
21
+ "MSE",
22
+ "PSNR",
23
+ "SSIM",
24
+ "LPIPS"
25
+ ],
26
+ "resolution_protocols": {
27
+ "fibo2-edit-turbo": "native-pad32-v1",
28
+ "joyai-image-edit-plus": "joyai-official-1024-bucket-resize-center-crop-v1",
29
+ "qwen-image21": "native-pad32-v1",
30
+ "qwen-image21-turbo": "native-pad32-v1",
31
+ "flux3-klein": "native-pad32-v1",
32
+ "sensenova-u1.5-8b-mot": "native-pad32-v1"
33
+ },
34
+ "model_configuration_notices": {
35
+ "joyai-image-edit-plus": "Official default bucket output,30steps,CFG4. Original source input without padding; omit width/height. Raw output is preserved. Pixel references/masks are mapped to the official output geometry. These pixel scores use a different resolution protocol from the other five models; report separately."
36
+ },
37
+ "created_at": "2026-11-08T14:42:22.846136+00:00",
38
+ "original_outputs": "Every raw_outputs PNG is byte-identical to the corresponding run3 model return.",
39
+ "portable_exports": "Use predictions_index.jsonl, not recorded_reports local-path metadata, to resolve files after download.",
40
+ "reports_note": "recorded_reports preserve historical reports with workspace root paths replaced by logical labels; they are reference records, not resumable caches."
41
+ }