-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathreal_e2e_example_run.html
More file actions
600 lines (600 loc) · 32 KB
/
real_e2e_example_run.html
File metadata and controls
600 lines (600 loc) · 32 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Real E2E Example Run ses_4c190e469983</title>
<style>
:root { color-scheme: light; font-family: Georgia, 'Times New Roman', serif; }
body { margin: 0; background: linear-gradient(180deg, #f1e8dc 0%, #fcfaf6 100%); color: #231b15; }
main { max-width: 1180px; margin: 0 auto; padding: 32px 20px 48px; }
.hero, .card { background: rgba(255, 252, 247, 0.96); border: 1px solid #d7ccbf; border-radius: 20px; padding: 24px; box-shadow: 0 14px 38px rgba(74, 50, 23, 0.08); }
.hero { margin-bottom: 18px; }
.eyebrow { text-transform: uppercase; letter-spacing: 0.12em; color: #87562a; font-size: 0.82rem; margin: 0 0 8px; }
h1, h2, h3 { margin-top: 0; font-family: 'Segoe UI', sans-serif; }
.lede { font-size: 1.08rem; line-height: 1.65; max-width: 900px; }
.metrics { display: grid; grid-template-columns: repeat(auto-fit, minmax(180px, 1fr)); gap: 12px; margin-top: 18px; }
.metric { background: #f8f1e7; border-radius: 16px; padding: 14px; }
.metric strong { display: block; font-family: 'Segoe UI', sans-serif; font-size: 0.82rem; text-transform: uppercase; letter-spacing: 0.08em; color: #765434; margin-bottom: 6px; }
.metric span { font-size: 1.02rem; font-weight: 700; }
.stack { display: grid; gap: 18px; }
.round { margin-top: 18px; }
.candidate-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 16px; }
.candidate { background: #fff; border-radius: 16px; overflow: hidden; border: 1px solid #e5d9cb; }
.candidate img { width: 100%; aspect-ratio: 1 / 1; object-fit: cover; display: block; background: #e9e0d4; }
.candidate .body { padding: 12px; font-size: 0.94rem; line-height: 1.45; }
.winner { outline: 3px solid #b16d1e; }
.pill { display: inline-block; margin-right: 8px; margin-bottom: 8px; padding: 6px 10px; background: #efe2cf; border-radius: 999px; font-size: 0.84rem; }
.callout { background: #f6efe4; border-left: 4px solid #a55b16; border-radius: 14px; padding: 16px 18px; }
.columns { display: grid; grid-template-columns: repeat(auto-fit, minmax(280px, 1fr)); gap: 16px; }
pre { white-space: pre-wrap; word-break: break-word; background: #f7f1e9; border-radius: 14px; padding: 14px; font-size: 0.88rem; }
table { width: 100%; border-collapse: collapse; }
th, td { text-align: left; vertical-align: top; padding: 10px; border-top: 1px solid #e7dbce; }
details.card { padding: 0; overflow: hidden; }
details.card > summary { list-style: none; cursor: pointer; padding: 22px 24px; font-family: 'Segoe UI', sans-serif; font-weight: 700; }
details.card > summary::-webkit-details-marker { display: none; }
details.card > summary::after { content: '+'; float: right; color: #87562a; }
details.card[open] > summary::after { content: '−'; }
.card-body { padding: 0 24px 24px; }
</style>
</head>
<body>
<main>
<section class="hero">
<p class="eyebrow">Real GPU Example</p>
<h1>Studio product photograph of a premium electric motorcycle, full side profile, isolated on a warm beige seamless backdrop, centered subject, catalog hero composition, polished body panels, no rider, no landscape</h1>
<p class="lede">This is a real end-to-end steering run executed against the Diffusers backend on GPU. It is designed to show the core value of the system: starting from a user text prompt, proposing multiple steerable directions, recording preference-guided choices, and moving the session toward a clearer creative target over multiple rounds.</p>
<div class="metrics">
<div class="metric"><strong>Session</strong><span>ses_4c190e469983</span></div>
<div class="metric"><strong>Backend</strong><span>diffusers</span></div>
<div class="metric"><strong>Device</strong><span>cuda</span></div>
<div class="metric"><strong>CUDA Available</strong><span>True</span></div>
<div class="metric"><strong>Rounds</strong><span>5</span></div>
<div class="metric"><strong>Model</strong><span>runwayml/stable-diffusion-v1-5</span></div>
<div class="metric"><strong>Unique images</strong><span>21</span></div>
</div>
</section>
<details class="card" open>
<summary>Demonstration Overview</summary>
<div class="card-body stack">
<div class="callout">
<h2>Demonstration Objective</h2>
<p>Demonstrate that StableSteering can start from a product-photography prompt and refine it over multiple rounds toward a cleaner, more premium, and more compositionally stable isolated hero image through iterative user preference rather than prompt rewriting.</p>
</div>
<div class="columns">
<div>
<h3>Success Criteria</h3>
<ul>
<li>The first round should include the unmodified-prompt baseline plus clearly different but still on-theme product-photo alternatives.</li>
<li>Later rounds should preserve the previous winner while still exploring nearby alternatives.</li>
<li>User preference should steer the session toward stronger side-profile silhouette, cleaner isolation, and more premium material detail.</li>
<li>By round five, the session should look more aligned with the desired hero-shot direction while preserving the overall product composition better than the initial round.</li>
<li>The report should make the proposal, selection, and update logic easy to inspect after the run.</li>
</ul>
</div>
<div>
<h3>Demo Script</h3>
<ol>
<li>Start from a user-written studio product prompt rather than a hidden preset.</li>
<li>Generate a first comparison round and inspect the baseline prompt render beside several steered studio-photo alternatives.</li>
<li>Choose the candidate with the strongest isolated side-profile framing and surface detail.</li>
<li>Generate additional rounds that keep the previous winner visible while exploring nearby refinements in geometry and background simplicity.</li>
<li>Continue for five rounds so the session history shows convergence instead of a one-step jump.</li>
<li>Review the report to see the proposed images, critiques, and incumbent transitions together.</li>
</ol>
</div>
</div>
</div>
</details>
<details class="card" open>
<summary>Run Setup</summary>
<div class="card-body stack">
<div>
<h2>Run Setup</h2>
<p><strong>Initial prompt:</strong> Studio product photograph of a premium electric motorcycle, full side profile, isolated on a warm beige seamless backdrop, centered subject, catalog hero composition, polished body panels, no rider, no landscape</p>
<p><strong>Negative prompt:</strong> person, rider, helmet, outdoors, desert, road, trees, sky, scenery, motion blur, blurry, low contrast, flat lighting, distorted wheels, text, watermark, cropped subject</p>
<p><strong>Sampler:</strong> <code>line_search</code> | <strong>Updater:</strong> <code>score_weighted_preference</code> | <strong>Feedback mode:</strong> <code>scalar_rating</code></p>
<p><strong>Output file:</strong> <code>E:\Projects\StableSteering\output\examples\real_e2e_example_run\real_e2e_example_run.html</code></p>
</div>
<pre>{
"active_device": "cuda",
"backend": "diffusers",
"configured_device": "cuda",
"cuda_available": true,
"default_model_source": "models\\runwayml--stable-diffusion-v1-5",
"default_num_inference_steps": 30,
"loaded_model_sources": [
"models\\runwayml--stable-diffusion-v1-5"
],
"local_files_only": true,
"model_source": "models\\runwayml--stable-diffusion-v1-5",
"pipeline_loaded": true,
"test_only_backend": false,
"torch_available": true
}</pre>
</div>
</details>
<details class="card" open>
<summary>Automated Visual Checks</summary>
<div class="card-body stack">
<div>
<h2>Automated Visual Checks</h2>
<p>These checks do not prove semantic correctness, but they do catch common generation failures such as blank-looking outputs, unreadable files, low-detail artifacts, and accidental duplication.</p>
</div>
<div class="metrics">
<div class="metric"><strong>Unique image files</strong><span>21</span></div>
<div class="metric"><strong>Reused candidates</strong><span>4</span></div>
<div class="metric"><strong>Flagged images</strong><span>0</span></div>
<div class="metric"><strong>Duplicate groups</strong><span>0</span></div>
</div>
<table>
<thead><tr><th>Image</th><th>Candidates</th><th>Size</th><th>Entropy</th><th>Contrast</th><th>Edges</th><th>Status</th></tr></thead>
<tbody>
<tr><td><code>images/cand_15d8f2add8fa.png</code></td><td>cand_15d8f2add8fa, cand_b1a8ad683e85</td><td>512x512</td><td>7.641</td><td>55.701</td><td>35.454</td><td>ok</td></tr>
<tr><td><code>images/cand_192e81f64493.png</code></td><td>cand_192e81f64493</td><td>512x512</td><td>7.926</td><td>69.23</td><td>22.91</td><td>ok</td></tr>
<tr><td><code>images/cand_21bd9a6129c7.png</code></td><td>cand_21bd9a6129c7</td><td>512x512</td><td>7.654</td><td>65.333</td><td>20.333</td><td>ok</td></tr>
<tr><td><code>images/cand_2672b5f21e74.png</code></td><td>cand_2672b5f21e74</td><td>512x512</td><td>7.391</td><td>62.025</td><td>21.209</td><td>ok</td></tr>
<tr><td><code>images/cand_27d807b768d1.png</code></td><td>cand_27d807b768d1</td><td>512x512</td><td>7.283</td><td>63.171</td><td>22.238</td><td>ok</td></tr>
<tr><td><code>images/cand_50cc80379f37.png</code></td><td>cand_50cc80379f37</td><td>512x512</td><td>7.823</td><td>69.114</td><td>18.852</td><td>ok</td></tr>
<tr><td><code>images/cand_5327b9b930b2.png</code></td><td>cand_5327b9b930b2</td><td>512x512</td><td>7.699</td><td>62.827</td><td>24.364</td><td>ok</td></tr>
<tr><td><code>images/cand_597d390ed370.png</code></td><td>cand_597d390ed370, cand_2aa7e07325e2</td><td>512x512</td><td>7.589</td><td>54.26</td><td>21.601</td><td>ok</td></tr>
<tr><td><code>images/cand_5b966aac15d1.png</code></td><td>cand_5b966aac15d1</td><td>512x512</td><td>7.659</td><td>54.573</td><td>35.951</td><td>ok</td></tr>
<tr><td><code>images/cand_5d9f24c47f68.png</code></td><td>cand_5d9f24c47f68</td><td>512x512</td><td>7.827</td><td>72.808</td><td>24.36</td><td>ok</td></tr>
<tr><td><code>images/cand_5e7e64d1e5db.png</code></td><td>cand_5e7e64d1e5db</td><td>512x512</td><td>7.658</td><td>64.238</td><td>37.189</td><td>ok</td></tr>
<tr><td><code>images/cand_623ae812fe7e.png</code></td><td>cand_623ae812fe7e</td><td>512x512</td><td>7.189</td><td>42.195</td><td>14.21</td><td>ok</td></tr>
<tr><td><code>images/cand_94f55a73cd1f.png</code></td><td>cand_94f55a73cd1f</td><td>512x512</td><td>7.488</td><td>59.806</td><td>21.449</td><td>ok</td></tr>
<tr><td><code>images/cand_a07e2cf2d77f.png</code></td><td>cand_a07e2cf2d77f</td><td>512x512</td><td>7.502</td><td>56.09</td><td>18.063</td><td>ok</td></tr>
<tr><td><code>images/cand_aa9f51e2210f.png</code></td><td>cand_aa9f51e2210f</td><td>512x512</td><td>7.163</td><td>43.215</td><td>19.88</td><td>ok</td></tr>
<tr><td><code>images/cand_bb6e17f5f09e.png</code></td><td>cand_bb6e17f5f09e</td><td>512x512</td><td>7.583</td><td>57.47</td><td>23.27</td><td>ok</td></tr>
<tr><td><code>images/cand_bd67b033d44d.png</code></td><td>cand_bd67b033d44d, cand_b646c1b57d4f, cand_c9237654bcb9</td><td>512x512</td><td>7.764</td><td>68.096</td><td>15.645</td><td>ok</td></tr>
<tr><td><code>images/cand_c1a969602297.png</code></td><td>cand_c1a969602297</td><td>512x512</td><td>7.817</td><td>63.562</td><td>30.321</td><td>ok</td></tr>
<tr><td><code>images/cand_dfb514537d2e.png</code></td><td>cand_dfb514537d2e</td><td>512x512</td><td>7.835</td><td>61.554</td><td>30.353</td><td>ok</td></tr>
<tr><td><code>images/cand_e89bb025c582.png</code></td><td>cand_e89bb025c582</td><td>512x512</td><td>7.464</td><td>53.92</td><td>19.635</td><td>ok</td></tr>
<tr><td><code>images/cand_f039d65ed06d.png</code></td><td>cand_f039d65ed06d</td><td>512x512</td><td>7.748</td><td>66.851</td><td>22.525</td><td>ok</td></tr>
</tbody>
</table>
<p>All automated visual sanity checks passed for the unique generated images in this bundle.</p>
</div>
</details>
<details class="card round">
<summary>Phase 1: propose, inspect, choose, update</summary>
<div class="card-body">
<p class="eyebrow">Round 1</p>
<p><span class="pill">Round id: rnd_585ebf7335f7</span><span class="pill">Latency: 75 ms</span><span class="pill">Incumbent z before update: [0.0, 0.0, 0.0, 0.0, 0.0]</span></p>
<div class="candidate-grid">
<article class="candidate">
<img src="images/cand_e89bb025c582.png" alt="cand_e89bb025c582">
<div class="body">
<p><strong>cand_e89bb025c582</strong></p>
<p>Role: baseline_prompt</p>
<p>Seed: 667071688</p>
<p>z: <code>[0.0, 0.0, 0.0, 0.0, 0.0]</code></p>
<p>User rating: <strong>3</strong></p>
</div>
</article>
<article class="candidate winner">
<img src="images/cand_bd67b033d44d.png" alt="cand_bd67b033d44d">
<div class="body">
<p><strong>cand_bd67b033d44d</strong></p>
<p>Role: forward_probe</p>
<p>Seed: 2650808045</p>
<p>z: <code>[0.5515076034546886, 0.29347004073709526, 0.14740359510060874, 0.10857196798139689, 8.025892956857929e-05]</code></p>
<p>User rating: <strong>5</strong></p>
<p><strong>Selected as winner for the next phase.</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_f039d65ed06d.png" alt="cand_f039d65ed06d">
<div class="body">
<p><strong>cand_f039d65ed06d</strong></p>
<p>Role: far_forward</p>
<p>Seed: 2479032873</p>
<p>z: <code>[0.09860170742472611, -0.5144148567432079, -0.29129822555236157, 0.16390908905678342, -0.12510050846827345]</code></p>
<p>User rating: <strong>4</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_5d9f24c47f68.png" alt="cand_5d9f24c47f68">
<div class="body">
<p><strong>cand_5d9f24c47f68</strong></p>
<p>Role: backtrack</p>
<p>Seed: 4236194812</p>
<p>z: <code>[0.06696506434237393, -0.011419845727026543, 0.5375779012963444, 0.29120578161753385, -0.17082154114604273]</code></p>
<p>User rating: <strong>2</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_5e7e64d1e5db.png" alt="cand_5e7e64d1e5db">
<div class="body">
<p><strong>cand_5e7e64d1e5db</strong></p>
<p>Role: lateral_probe</p>
<p>Seed: 4267972134</p>
<p>z: <code>[-0.1931886378426342, -0.054836379317133394, 0.002199341094889072, -0.5395699612994985, 0.30386105619822523]</code></p>
<p>User rating: <strong>1</strong></p>
</div>
</article>
</div>
<h3>User Action</h3>
<p>The baseline already points in the right product direction, but candidate 2 is the strongest isolated side-profile render with the cleanest silhouette, the most centered framing, and the least distracting scene content. Candidate 3 is also plausible, but it is slightly weaker in product readability.</p>
<pre>{
"ratings": {
"cand_5d9f24c47f68": 2,
"cand_5e7e64d1e5db": 1,
"cand_bd67b033d44d": 5,
"cand_e89bb025c582": 3,
"cand_f039d65ed06d": 4
},
"winner_candidate_id": "cand_bd67b033d44d"
}</pre>
<h3>Next Phase State</h3>
<pre>{
"method": "score_weighted_centroid",
"updater": "score_weighted_preference",
"weight_count": 5,
"weights": {
"cand_5d9f24c47f68": 2.0,
"cand_5e7e64d1e5db": 1.0,
"cand_bd67b033d44d": 5.0,
"cand_e89bb025c582": 3.0,
"cand_f039d65ed06d": 4.0
},
"winner_candidate_id": "cand_bd67b033d44d"
}</pre>
</div>
</details>
<details class="card round">
<summary>Phase 2: propose, inspect, choose, update</summary>
<div class="card-body">
<p class="eyebrow">Round 2</p>
<p><span class="pill">Round id: rnd_1ab75ade2525</span><span class="pill">Latency: 75 ms</span><span class="pill">Incumbent z before update: [0.140202, -0.030282, 0.02943, 0.056274, -0.024379]</span></p>
<div class="candidate-grid">
<article class="candidate winner">
<img src="images/cand_bd67b033d44d.png" alt="cand_b646c1b57d4f">
<div class="body">
<p><strong>cand_b646c1b57d4f</strong></p>
<p>Role: incumbent</p>
<p>Seed: 2650808045</p>
<p>z: <code>[0.5515076034546886, 0.29347004073709526, 0.14740359510060874, 0.10857196798139689, 8.025892956857929e-05]</code></p>
<p>User rating: <strong>5</strong></p>
<p><strong>Selected as winner for the next phase.</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_27d807b768d1.png" alt="cand_27d807b768d1">
<div class="body">
<p><strong>cand_27d807b768d1</strong></p>
<p>Role: forward_probe</p>
<p>Seed: 3090523580</p>
<p>z: <code>[0.2699873157178235, -0.03761694504120552, 0.03856881054146627, 0.11441318649490592, -0.06023058163450399]</code></p>
<p>User rating: <strong>4</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_94f55a73cd1f.png" alt="cand_94f55a73cd1f">
<div class="body">
<p><strong>cand_94f55a73cd1f</strong></p>
<p>Role: far_forward</p>
<p>Seed: 3242312771</p>
<p>z: <code>[0.36594503913303406, -0.09558450059627496, 0.07780958041531204, 0.15537376026050503, -0.05614708717099261]</code></p>
<p>User rating: <strong>2</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_50cc80379f37.png" alt="cand_50cc80379f37">
<div class="body">
<p><strong>cand_50cc80379f37</strong></p>
<p>Role: backtrack</p>
<p>Seed: 921904678</p>
<p>z: <code>[0.03512890284314759, -0.019624006683121135, -0.007540652859610138, 0.013062115849022185, 0.004740116709586201]</code></p>
<p>User rating: <strong>3</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_aa9f51e2210f.png" alt="cand_aa9f51e2210f">
<div class="body">
<p><strong>cand_aa9f51e2210f</strong></p>
<p>Role: lateral_probe</p>
<p>Seed: 3329634114</p>
<p>z: <code>[0.20933601100029608, 0.19207936750180243, 0.03577742088218002, 0.11517696643708371, -0.025322152712335944]</code></p>
<p>User rating: <strong>1</strong></p>
</div>
</article>
</div>
<h3>User Action</h3>
<p>Keep the incumbent, but refine toward cleaner wheel geometry, a calmer seamless backdrop, and a more catalog-like isolated composition.</p>
<pre>{
"ratings": {
"cand_27d807b768d1": 4,
"cand_50cc80379f37": 3,
"cand_94f55a73cd1f": 2,
"cand_aa9f51e2210f": 1,
"cand_b646c1b57d4f": 5
},
"winner_candidate_id": "cand_b646c1b57d4f"
}</pre>
<h3>Next Phase State</h3>
<pre>{
"method": "score_weighted_centroid",
"updater": "score_weighted_preference",
"weight_count": 5,
"weights": {
"cand_27d807b768d1": 4.0,
"cand_50cc80379f37": 3.0,
"cand_94f55a73cd1f": 2.0,
"cand_aa9f51e2210f": 1.0,
"cand_b646c1b57d4f": 5.0
},
"winner_candidate_id": "cand_b646c1b57d4f"
}</pre>
</div>
</details>
<details class="card round">
<summary>Phase 3: propose, inspect, choose, update</summary>
<div class="card-body">
<p class="eyebrow">Round 3</p>
<p><span class="pill">Round id: rnd_28d0cd0a34d6</span><span class="pill">Latency: 75 ms</span><span class="pill">Incumbent z before update: [0.266277, 0.047381, 0.057474, 0.084449, -0.024299]</span></p>
<div class="candidate-grid">
<article class="candidate">
<img src="images/cand_bd67b033d44d.png" alt="cand_c9237654bcb9">
<div class="body">
<p><strong>cand_c9237654bcb9</strong></p>
<p>Role: incumbent</p>
<p>Seed: 2650808045</p>
<p>z: <code>[0.5515076034546886, 0.29347004073709526, 0.14740359510060874, 0.10857196798139689, 8.025892956857929e-05]</code></p>
<p>User rating: <strong>4</strong></p>
</div>
</article>
<article class="candidate winner">
<img src="images/cand_15d8f2add8fa.png" alt="cand_15d8f2add8fa">
<div class="body">
<p><strong>cand_15d8f2add8fa</strong></p>
<p>Role: forward_probe</p>
<p>Seed: 429822218</p>
<p>z: <code>[0.38046676423133885, 0.06754211814611827, 0.09569082989925647, 0.13205726189940017, -0.02207370029005937]</code></p>
<p>User rating: <strong>5</strong></p>
<p><strong>Selected as winner for the next phase.</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_5b966aac15d1.png" alt="cand_5b966aac15d1">
<div class="body">
<p><strong>cand_5b966aac15d1</strong></p>
<p>Role: far_forward</p>
<p>Seed: 2144882815</p>
<p>z: <code>[0.3930850015313224, 0.05184587533897571, 0.07021609398282626, 0.1172721041943535, -0.022652604818854404]</code></p>
<p>User rating: <strong>3</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_5327b9b930b2.png" alt="cand_5327b9b930b2">
<div class="body">
<p><strong>cand_5327b9b930b2</strong></p>
<p>Role: backtrack</p>
<p>Seed: 3621839002</p>
<p>z: <code>[0.1400826975662036, 0.03642577725181688, 0.011614538510324384, 0.04915016209197056, -0.01428621503671396]</code></p>
<p>User rating: <strong>2</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_a07e2cf2d77f.png" alt="cand_a07e2cf2d77f">
<div class="body">
<p><strong>cand_a07e2cf2d77f</strong></p>
<p>Role: lateral_probe</p>
<p>Seed: 1703667268</p>
<p>z: <code>[0.2242195663361524, 0.2524151538291431, 0.03809549240435736, 0.12893629380305024, -0.026739690983646036]</code></p>
<p>User rating: <strong>1</strong></p>
</div>
</article>
</div>
<h3>User Action</h3>
<p>Candidate 2 improves the overall silhouette and wheel balance while preserving the centered isolated product-shot framing.</p>
<pre>{
"ratings": {
"cand_15d8f2add8fa": 5,
"cand_5327b9b930b2": 2,
"cand_5b966aac15d1": 3,
"cand_a07e2cf2d77f": 1,
"cand_c9237654bcb9": 4
},
"winner_candidate_id": "cand_15d8f2add8fa"
}</pre>
<h3>Next Phase State</h3>
<pre>{
"method": "score_weighted_centroid",
"updater": "score_weighted_preference",
"weight_count": 5,
"weights": {
"cand_15d8f2add8fa": 5.0,
"cand_5327b9b930b2": 2.0,
"cand_5b966aac15d1": 3.0,
"cand_a07e2cf2d77f": 1.0,
"cand_c9237654bcb9": 4.0
},
"winner_candidate_id": "cand_15d8f2add8fa"
}</pre>
</div>
</details>
<details class="card round">
<summary>Phase 4: propose, inspect, choose, update</summary>
<div class="card-body">
<p class="eyebrow">Round 4</p>
<p><span class="pill">Round id: rnd_d1b7b8cfe89e</span><span class="pill">Latency: 75 ms</span><span class="pill">Incumbent z before update: [0.347779, 0.105484, 0.07914, 0.102895, -0.018353]</span></p>
<div class="candidate-grid">
<article class="candidate">
<img src="images/cand_15d8f2add8fa.png" alt="cand_b1a8ad683e85">
<div class="body">
<p><strong>cand_b1a8ad683e85</strong></p>
<p>Role: incumbent</p>
<p>Seed: 429822218</p>
<p>z: <code>[0.38046676423133885, 0.06754211814611827, 0.09569082989925647, 0.13205726189940017, -0.02207370029005937]</code></p>
<p>User rating: <strong>4</strong></p>
</div>
</article>
<article class="candidate winner">
<img src="images/cand_597d390ed370.png" alt="cand_597d390ed370">
<div class="body">
<p><strong>cand_597d390ed370</strong></p>
<p>Role: forward_probe</p>
<p>Seed: 4153441364</p>
<p>z: <code>[0.3807590858494705, 0.10301919077091298, 0.09836939210259654, 0.10494938640261402, -0.010892835584101075]</code></p>
<p>User rating: <strong>5</strong></p>
<p><strong>Selected as winner for the next phase.</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_c1a969602297.png" alt="cand_c1a969602297">
<div class="body">
<p><strong>cand_c1a969602297</strong></p>
<p>Role: far_forward</p>
<p>Seed: 2121964274</p>
<p>z: <code>[0.3765601658470825, 0.11218968248140983, 0.07469262865776699, 0.12731859229892553, -0.015063330347272507]</code></p>
<p>User rating: <strong>3</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_623ae812fe7e.png" alt="cand_623ae812fe7e">
<div class="body">
<p><strong>cand_623ae812fe7e</strong></p>
<p>Role: backtrack</p>
<p>Seed: 2978702666</p>
<p>z: <code>[0.20423130996596994, 0.04881720572883034, 0.05668903684246285, 0.056990529964534756, -0.023094568408170972]</code></p>
<p>User rating: <strong>2</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_bb6e17f5f09e.png" alt="cand_bb6e17f5f09e">
<div class="body">
<p><strong>cand_bb6e17f5f09e</strong></p>
<p>Role: lateral_probe</p>
<p>Seed: 1922418150</p>
<p>z: <code>[0.26525818047729754, 0.2938992802062353, 0.04945744028587103, 0.12500696754292367, -0.004236409687446114]</code></p>
<p>User rating: <strong>1</strong></p>
</div>
</article>
</div>
<h3>User Action</h3>
<p>Candidate 2 has the best balance of realistic product geometry, clean background control, and a readable side-profile silhouette. Keep that direction.</p>
<pre>{
"ratings": {
"cand_597d390ed370": 5,
"cand_623ae812fe7e": 2,
"cand_b1a8ad683e85": 4,
"cand_bb6e17f5f09e": 1,
"cand_c1a969602297": 3
},
"winner_candidate_id": "cand_597d390ed370"
}</pre>
<h3>Next Phase State</h3>
<pre>{
"method": "score_weighted_centroid",
"updater": "score_weighted_preference",
"weight_count": 5,
"weights": {
"cand_597d390ed370": 5.0,
"cand_623ae812fe7e": 2.0,
"cand_b1a8ad683e85": 4.0,
"cand_bb6e17f5f09e": 1.0,
"cand_c1a969602297": 3.0
},
"winner_candidate_id": "cand_597d390ed370"
}</pre>
</div>
</details>
<details class="card round">
<summary>Phase 5: propose, inspect, choose, update</summary>
<div class="card-body">
<p class="eyebrow">Round 5</p>
<p><span class="pill">Round id: rnd_5904bd426422</span><span class="pill">Latency: 75 ms</span><span class="pill">Incumbent z before update: [0.34834, 0.102361, 0.082514, 0.108811, -0.016679]</span></p>
<div class="candidate-grid">
<article class="candidate winner">
<img src="images/cand_597d390ed370.png" alt="cand_2aa7e07325e2">
<div class="body">
<p><strong>cand_2aa7e07325e2</strong></p>
<p>Role: incumbent</p>
<p>Seed: 4153441364</p>
<p>z: <code>[0.3807590858494705, 0.10301919077091298, 0.09836939210259654, 0.10494938640261402, -0.010892835584101075]</code></p>
<p>User rating: <strong>5</strong></p>
<p><strong>Selected as winner for the next phase.</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_192e81f64493.png" alt="cand_192e81f64493">
<div class="body">
<p><strong>cand_192e81f64493</strong></p>
<p>Role: forward_probe</p>
<p>Seed: 2041570376</p>
<p>z: <code>[0.373727585109508, 0.11036075523548379, 0.087842219284045, 0.12970179976240176, -0.003063246923890366]</code></p>
<p>User rating: <strong>4</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_2672b5f21e74.png" alt="cand_2672b5f21e74">
<div class="body">
<p><strong>cand_2672b5f21e74</strong></p>
<p>Role: far_forward</p>
<p>Seed: 1936755664</p>
<p>z: <code>[0.3847011810046245, 0.09500887061369813, 0.07816807946806743, 0.11268493452808452, -0.023878289742606425]</code></p>
<p>User rating: <strong>3</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_21bd9a6129c7.png" alt="cand_21bd9a6129c7">
<div class="body">
<p><strong>cand_21bd9a6129c7</strong></p>
<p>Role: backtrack</p>
<p>Seed: 485589457</p>
<p>z: <code>[0.22489770375721055, 0.07895322630755397, 0.07713999111647796, 0.08696927911724772, 0.002124418872091983]</code></p>
<p>User rating: <strong>2</strong></p>
</div>
</article>
<article class="candidate">
<img src="images/cand_dfb514537d2e.png" alt="cand_dfb514537d2e">
<div class="body">
<p><strong>cand_dfb514537d2e</strong></p>
<p>Role: lateral_probe</p>
<p>Seed: 1511722177</p>
<p>z: <code>[0.26553313020842423, 0.2969832771925788, 0.06594769547223969, 0.11532026564668224, -0.006725148530053199]</code></p>
<p>User rating: <strong>1</strong></p>
</div>
</article>
</div>
<h3>User Action</h3>
<p>The incumbent now reads like the strongest finished isolated hero image, so preserve it as the final preferred direction.</p>
<pre>{
"ratings": {
"cand_192e81f64493": 4,
"cand_21bd9a6129c7": 2,
"cand_2672b5f21e74": 3,
"cand_2aa7e07325e2": 5,
"cand_dfb514537d2e": 1
},
"winner_candidate_id": "cand_2aa7e07325e2"
}</pre>
<h3>Next Phase State</h3>
<pre>{
"method": "score_weighted_centroid",
"updater": "score_weighted_preference",
"weight_count": 5,
"weights": {
"cand_192e81f64493": 4.0,
"cand_21bd9a6129c7": 2.0,
"cand_2672b5f21e74": 3.0,
"cand_2aa7e07325e2": 5.0,
"cand_dfb514537d2e": 1.0
},
"winner_candidate_id": "cand_2aa7e07325e2"
}</pre>
</div>
</details>
<details class="card" open>
<summary>Outcome</summary>
<div class="card-body">
<h2>Outcome</h2>
<p>The session ended at round <strong>5</strong> with incumbent candidate <code>cand_2aa7e07325e2</code>.</p>
<p>This report demonstrates system value by showing that the workflow does not stop at one prompt and one image. It keeps the user in a controlled compare-select-update loop with visible evidence for what was proposed, what was preferred, and how the next state was chosen.</p>
<p>Backend session traces and the auto-generated audit report also exist under <code>E:\Projects\StableSteering\output\examples\real_e2e_example_run\data\traces</code>.</p>
</div>
</details>
</main>
</body>
</html>