{ "disclaimer_en": "Informal community evaluation for reference only; not an official evaluation.", "disclaimer_zh": "社区非正式测评,仅供参考,不代表任何官方评价。", "cases": [ { "id": "sweater_red", "en": "Older woman: blue sweater to red", "zh": "老年女性:蓝色毛衣改为红色", "prompt": "Change only the blue sweater to a red sweater. Preserve the same person, face, pose, lighting and background.", "input": "evaluation/audit/input.png", "bf16": "evaluation/audit/bf16-native-tiled/edit_native.png", "int8": "evaluation/audit/int8-native-tiled/edit_native.png", "observation_en": "Both outputs make the sweater red and retain soft lighting and broadly natural skin texture. Fine facial and fabric details still shift, and faint colored marks are visible in both; this is not lossless preservation.", "observation_zh": "两组均将毛衣改为红色,较好保留柔和光线与自然皮肤质感;面部和织物细节仍有偏移,两组均可见细小彩色痕迹,并非无损保持。", "input_sha256": "01e37e178fb0d35d5c706b7edd76bc1a6c5b50ec3536901a467fd45295d097aa", "bf16_sha256": "c551ac809ccb48146d1ad842d75975c0e9663d388a368010d2bfccb7883eeeec", "int8_sha256": "b6af18d8688bb2669632adf0cfae962bc34431e30e935014997010cde5c2ca59" }, { "id": "portrait_recolor", "en": "Adult woman: cream sweater to navy blue", "zh": "成年女性:米白毛衣改为藏蓝", "prompt": "Change only the cream sweater to a navy blue sweater. Preserve the same woman, face, hairstyle, pose, lighting and background.", "input": "evaluation/editing/inputs/portrait_recolor.png", "bf16": "evaluation/editing/bf16/portrait_recolor-s42.png", "int8": "evaluation/editing/int8/portrait_recolor-s42.png", "observation_en": "Both outputs recolor the sweater to navy blue while broadly retaining the woman, pose, window light and background. Facial, hair and knitted-fabric details still shift slightly; this is not an identity-preservation score.", "observation_zh": "两组均将毛衣改为藏蓝色,大体保持人物、姿态、窗边光线与背景。五官、发丝和针织纹理仍有轻微变化,这不是身份保持评分。", "input_sha256": "00716ad4721f585be4b0bf0daf2413f4ab92b795f1eb4d08b4f1ecc604b075b8", "bf16_sha256": "a215c639d838aa9080346a6701008ce991e90f278db175ec7f110bf244f19e7c", "int8_sha256": "d9d566f33d7d8c82b65b009766212d2b68d0f34e1b9a42e538af979b59c1bda4" }, { "id": "cup_material", "en": "Product edit: ceramic cup to brushed metal", "zh": "物品编辑:陶瓷杯改为拉丝金属", "prompt": "Change only the middle blue ceramic cup into a brushed silver metal cup. Preserve its shape and position, the red and yellow cups, the green apple, the tabletop, lighting and background.", "input": "evaluation/editing/inputs/cup_material.png", "bf16": "evaluation/editing/bf16/cup_material-s42.png", "int8": "evaluation/editing/int8/cup_material-s42.png", "observation_en": "Both make the middle cup look like brushed silver metal while broadly preserving the arrangement of three cups and the apple. Cup rims, reflections and apple texture shift slightly; faint line artifacts are visible in the background.", "observation_zh": "两组均将中间杯子呈现为银色拉丝金属外观,大体保持三杯与苹果的排列。杯口、反光和苹果纹理有轻微变化,背景可见细线状痕迹。", "input_sha256": "91e967685fbbd79ba13daece22a0f72c863cb518db88d5d2a6f9aecfc5f292b1", "bf16_sha256": "4c5f4f22843ca58ff8f52ca5f5b62061083b746dc3f14cb0a1827bcfcdc9df6e", "int8_sha256": "9177ea912de8d1c6e7c99782d939acbaaa0049da14784b4b5fdb81b4390d4ce0" }, { "id": "bird_background", "en": "Background replacement: snowy forest (unsuccessful)", "zh": "背景替换:雪林(未达到目标)", "prompt": "Replace only the blurred forest background with a softly blurred snowy winter forest. Preserve the same kingfisher, its colors and pose, the mossy branch, framing and foreground lighting.", "input": "evaluation/editing/inputs/bird_background.png", "bf16": "evaluation/editing/bf16/bird_background-s42.png", "int8": "evaluation/editing/int8/bird_background-s42.png", "observation_en": "Both fail the requested edit: the mossy branch disappears, the background becomes largely white instead of a snowy forest, and the bird loses photographic detail. This failure is present in both BF16 and INT8 and cannot be attributed to quantization alone.", "observation_zh": "两组均未达到编辑目标:苔藓枝干消失,背景大面积变白而非雪林,鸟的摄影细节也明显损失。BF16 与 INT8 均存在此失败,不能将其单独归因于量化。", "input_sha256": "40a0f73bb407a394fa9e8a3f6922455df716b9725ad07437fa10c1daca2c2095", "bf16_sha256": "040480b513d46eadb0d13910845d1f539fad7152dd57375892bb574969430ef8", "int8_sha256": "7afde64515861b1312388f88edf94c25dc146b388107cd223befafc3e22e3eac" } ], "measurements": [ { "case_id": "portrait_recolor", "bf16_seconds": 201.80096119999507, "int8_seconds": 194.7663809000005, "bf16_peak_gib": 26.51371479034424, "int8_peak_gib": 21.646138191223145 }, { "case_id": "cup_material", "bf16_seconds": 433.98883840000053, "int8_seconds": 193.33272730000317, "bf16_peak_gib": 26.52234125137329, "int8_peak_gib": 21.656850337982178 }, { "case_id": "bird_background", "bf16_seconds": 585.3128182, "int8_seconds": 184.6775536999994, "bf16_peak_gib": 26.521997451782227, "int8_peak_gib": 21.65667486190796 } ], "case_file_note": "Published JSON text may use LF line endings. measured-cases.zip preserves the exact editing.json bytes whose SHA256 appears in both environment records; parsed case definitions are identical.", "timing_note": "Three declared cases, one excluded seed-0 warmup per precision, one measured seed-42 call per case. The separate older-woman cold instrumented call is excluded." }