id string | dataset_row_id string | sample_row_id string | sample_id int64 | split string | strategy string | strategy_description string | experiment_type string | context_count int64 | package string | chunk string | relative_path string | system_prompt string | judge_prompt string | messages list | prompt_images images list | context_image_indices list | candidate_image_indices list | image_inputs list | candidate_order list | ground_truth_candidate_index int64 | ground_truth_candidate_label string | ground_truth_candidate_kind string | judge_model string | judge_url string | judge_temperature float64 | judge_top_p float64 | judge_top_k int64 | judge_num_predict int64 | judge_crop_original_top_px int64 | judge_crop_context_top_px int64 | judge_fit_rendered_to_original bool | judge_rendered_target_size_px list | judge_candidate_order_seed int64 | judge_rollout_seed int64 | judge_rollout_count int64 | source dict |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
anvil-17633648-sample000001 | anvil-17633648-sample000001 | anvil-17633648-sample000001 | 1 | train | cross-3 | cross-strategy (3 refs, random) | design_system_match | 3 | com.quora.android | com.quora.android__chunk_0000 | com.quora.android/com.quora.android__chunk_0000/screenshots/2024_4_30_12_58_8ac5251dfec547e29851f6e8132aa9d8-10.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2
] | [
3,
4
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000001/context-00.png",
"sha256": "5f0bacb0a07b139eeb2aa7c68cdb0998bb08a6ec078c44b1b2dbb765a12019e1",
"size_bytes": 1550018,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 3,
"source_path": "assets/anvil-17633648/sample-000001/original.png",
"sha256": "3d0c9d9b3e61519cc037aa9fe5c043ed1700e5731b94c01279bd234a204a87b4",
"size_bytes": 1491086,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 239,081,665 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 1,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000001",
"source_dataset_row_id": "anvil-17633648-sample000001-rollout01",
"source_split": "train",
"source_row_numbers": [
1,
2,
3,
4,
... | |
anvil-17633648-sample000003 | anvil-17633648-sample000003 | anvil-17633648-sample000003 | 3 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.amazon.mShop.android.shopping | com.amazon.mShop.android.shopping__chunk_0000 | com.amazon.mShop.android.shopping/com.amazon.mShop.android.shopping__chunk_0000/screenshots/2024_5_14_14_3_548aaf55da274062a44432100d6549fc-12.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "original",
"source_path": "assets/anvil-17633648/sample-000003/original.png",
"sha256": "92374f7ee79fcf7948496da226c91e5784a81760fe409ed105cc00d809277911",
"size_bytes": 1509361,
"crop_top": 0,
"... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000003/original.png",
"sha256": "92374f7ee79fcf7948496da226c91e5784a81760fe409ed105cc00d809277911",
"size_bytes": 1509361,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 525,901,260 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 2,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000003",
"source_dataset_row_id": "anvil-17633648-sample000003-rollout01",
"source_split": "train",
"source_row_numbers": [
9,
10,
11,
12,
... | |
anvil-17633648-sample000005 | anvil-17633648-sample000005 | anvil-17633648-sample000005 | 5 | train | cross-5 | cross-strategy (5 refs, random) | design_system_match | 5 | com.aspiro.tidal | com.aspiro.tidal__chunk_0000 | com.aspiro.tidal/com.aspiro.tidal__chunk_0000/screenshots/2024_4_25_11_38_1b590bfe8e4b4d1a9a9aac137e9b5546-3.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4
] | [
5,
6
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000005/context-00.png",
"sha256": "d70f3edc2950f0a643170261b1bc8baeb21ee315213f9eb65138cb312887f521",
"size_bytes": 345156,
"crop_top": 110,
"fit_... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 5,
"source_path": "assets/anvil-17633648/sample-000005/original.png",
"sha256": "27a9c71c3001c119d3a7b3ffd1b1af740b78b3c1b2907b8d247198efdc15741e",
"size_bytes": 249493,
"crop_top": 110,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 110 | false | [] | 220,106,713 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 4,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000005",
"source_dataset_row_id": "anvil-17633648-sample000005-rollout01",
"source_split": "train",
"source_row_numbers": [
25,
26,
27,
28,... | |
anvil-17633648-sample000006 | anvil-17633648-sample000006 | anvil-17633648-sample000006 | 6 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.bambuna.podcastaddict | com.bambuna.podcastaddict__chunk_0001 | com.bambuna.podcastaddict/com.bambuna.podcastaddict__chunk_0001/screenshots/2024_4_22_16_23_90878676290944348328790546366fbf-8.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000006/context-00.png",
"sha256": "ac7c1be4b3d08491268b393795e5996545c420d8cdcf0b467cc54c080ca9a6d2",
"size_bytes": 1541539,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000006/original.png",
"sha256": "3533c058a9da3e4871ed37950041d739de377b306b747de54f75c7df4ec45431",
"size_bytes": 2265310,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 1,590,571,872 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 5,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000006",
"source_dataset_row_id": "anvil-17633648-sample000006-rollout01",
"source_split": "train",
"source_row_numbers": [
33,
34,
35,
36,... | |
anvil-17633648-sample000009 | anvil-17633648-sample000009 | anvil-17633648-sample000009 | 9 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.yelp.android | com.yelp.android__chunk_0000 | com.yelp.android/com.yelp.android__chunk_0000/screenshots/2024_6_19_17_50_8e50dda246434698ad00b5fe62f2378f-10.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000009/context-00.png",
"sha256": "0824f0f7902afc3577132784f1034d1fd487d9f997c9f9f5b35a807b31d8581b",
"size_bytes": 469552,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000009/rendered.png",
"sha256": "8e283902d4ae5fb08b5980d83e01314d625bde7b8627a890bdb03eb18be0496f",
"size_bytes": 3310487,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 68,252,803 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 8,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000009",
"source_dataset_row_id": "anvil-17633648-sample000009-rollout01",
"source_split": "train",
"source_row_numbers": [
57,
58,
59,
60,... | |
anvil-17633648-sample000010 | anvil-17633648-sample000010 | anvil-17633648-sample000010 | 10 | train | cross-3 | cross-strategy (3 refs, random) | design_system_match | 3 | bbc.mobile.sport.ww | bbc.mobile.sport.ww__chunk_0000 | bbc.mobile.sport.ww/bbc.mobile.sport.ww__chunk_0000/screenshots/2024_4_25_22_3_da58e5168c5948a4b4518ef679d27c15-9.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2
] | [
3,
4
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000010/context-00.png",
"sha256": "0a6a16d19d1239a8eff8e1c0af92433bc3b1cfa98bfa8a491c9a7ef5bd86bd16",
"size_bytes": 497510,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 3,
"source_path": "assets/anvil-17633648/sample-000010/rendered.png",
"sha256": "0bcca4b713ef5916cdd1308441be66c033cb9e8930f0a5e391eda9e51acf0401",
"size_bytes": 278304,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 201,209,016 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 9,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000010",
"source_dataset_row_id": "anvil-17633648-sample000010-rollout01",
"source_split": "train",
"source_row_numbers": [
65,
66,
67,
68,... | |
anvil-17633648-sample000011 | anvil-17633648-sample000011 | anvil-17633648-sample000011 | 11 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.google.android.gm | com.google.android.gm__chunk_0000 | com.google.android.gm/com.google.android.gm__chunk_0000/screenshots/2024_5_6_20_22_3052b8a651044a129575be958a91b3cd-10.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000011/context-00.png",
"sha256": "38ff5edf044ca2573cbfc5062d314751dd107697e27bf0dca2bb22d2745f412b",
"size_bytes": 561629,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000011/original.png",
"sha256": "bac9e6ad66b7bd744c9d56071c48d0387f575597d9eb469924945b5b302b4275",
"size_bytes": 543291,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 499,635,480 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 10,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000011",
"source_dataset_row_id": "anvil-17633648-sample000011-rollout01",
"source_split": "train",
"source_row_numbers": [
73,
74,
75,
76... | |
anvil-17633648-sample000015 | anvil-17633648-sample000015 | anvil-17633648-sample000015 | 15 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.google.android.gm | com.google.android.gm__chunk_0000 | com.google.android.gm/com.google.android.gm__chunk_0000/screenshots/2024_5_6_19_53_f76dcb4f03eb4e27a4f490bd85e6e6b5-8.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "original",
"source_path": "assets/anvil-17633648/sample-000015/original.png",
"sha256": "a9a3d6b071cd718799e4db4d6aad0c828e286f1b4d32c5460b6db35189a96897",
"size_bytes": 850761,
"crop_top": 0,
"f... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000015/original.png",
"sha256": "a9a3d6b071cd718799e4db4d6aad0c828e286f1b4d32c5460b6db35189a96897",
"size_bytes": 850761,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 900,911,970 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 13,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000015",
"source_dataset_row_id": "anvil-17633648-sample000015-rollout01",
"source_split": "train",
"source_row_numbers": [
97,
98,
99,
10... | |
anvil-17633648-sample000016 | anvil-17633648-sample000016 | anvil-17633648-sample000016 | 16 | train | cross-3 | cross-strategy (3 refs, random) | design_system_match | 3 | com.zzkko | com.zzkko__chunk_0002 | com.zzkko/com.zzkko__chunk_0002/screenshots/2024_5_6_14_46_ddfe71c0ae264b989d1a51ddda727d6c-14.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2
] | [
3,
4
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000016/context-00.png",
"sha256": "c44ca950f252cf2b80bacf8159dd979de33b5988cfdb78b1eeeff5f9c7555e79",
"size_bytes": 1459635,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 3,
"source_path": "assets/anvil-17633648/sample-000016/rendered.png",
"sha256": "9dba5e8abf7bf35c3deea09140cf3ef9f335bc14e934b1aa8a9904890755bdcc",
"size_bytes": 1199307,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 964,669,094 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 14,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000016",
"source_dataset_row_id": "anvil-17633648-sample000016-rollout01",
"source_split": "train",
"source_row_numbers": [
105,
106,
107,
... | |
anvil-17633648-sample000018 | anvil-17633648-sample000018 | anvil-17633648-sample000018 | 18 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.nbaimd.gametime.nba2011 | com.nbaimd.gametime.nba2011__chunk_0000 | com.nbaimd.gametime.nba2011/com.nbaimd.gametime.nba2011__chunk_0000/screenshots/2024_4_1_19_14_6f9177f306da4713acf8566cd752a641-7.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000018/context-00.png",
"sha256": "ba68395ff105736a928f9fa6f838b4f7fbbe4beb3de812c4083924c6ed2e0eb3",
"size_bytes": 3319612,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000018/rendered.png",
"sha256": "0d1ce9960f39e3cdc9bf523a4cc1ee20dafb2e2e941da0b1d0f459bded28e46a",
"size_bytes": 387731,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 1,866,808,248 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 16,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000018",
"source_dataset_row_id": "anvil-17633648-sample000018-rollout01",
"source_split": "train",
"source_row_numbers": [
121,
122,
123,
... | |
anvil-17633648-sample000020 | anvil-17633648-sample000020 | anvil-17633648-sample000020 | 20 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.tencent.ibg.joox | com.tencent.ibg.joox__chunk_0000 | com.tencent.ibg.joox/com.tencent.ibg.joox__chunk_0000/screenshots/2024_3_22_22_55_466a682972564ed293c2f786e4446854-7.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000020/context-00.png",
"sha256": "86e4a2408e5ce8e06e8c04689f57c577af22f085f507e8c9e41f7740fe43c4f5",
"size_bytes": 318326,
"crop_top": 68,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000020/original.png",
"sha256": "8850d04033c6a10ddb8df3032dcd9542f9861ecebe254c879ad09824a0c2da44",
"size_bytes": 834655,
"crop_top": 68,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 68 | false | [] | 342,865,783 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 17,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000020",
"source_dataset_row_id": "anvil-17633648-sample000020-rollout01",
"source_split": "train",
"source_row_numbers": [
129,
130,
131,
... | |
anvil-17633648-sample000022 | anvil-17633648-sample000022 | anvil-17633648-sample000022 | 22 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.microsoft.teams | com.microsoft.teams__chunk_0000 | com.microsoft.teams/com.microsoft.teams__chunk_0000/screenshots/2024_4_24_21_5_1af3c14e962047b6b8c7c9d9a5dfca74-3.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000022/context-00.png",
"sha256": "9ff9032a34d7025a6ab159d9ad9ebace4ea18de586acf2374e1bf60d78e14b03",
"size_bytes": 470076,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000022/rendered.png",
"sha256": "992bc0bacf23c5b85a5eea98f6126ea8e0b9f976835e0e315e8af51139ed40a2",
"size_bytes": 646229,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 462,382,804 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 18,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000022",
"source_dataset_row_id": "anvil-17633648-sample000022-rollout01",
"source_split": "train",
"source_row_numbers": [
137,
138,
139,
... | |
anvil-17633648-sample000023 | anvil-17633648-sample000023 | anvil-17633648-sample000023 | 23 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | bbc.mobile.sport.ww | bbc.mobile.sport.ww__chunk_0000 | bbc.mobile.sport.ww/bbc.mobile.sport.ww__chunk_0000/screenshots/2024_4_25_14_6_80ab8a1eb7d44d50b9c58dfce782ffb3-8.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "original",
"source_path": "assets/anvil-17633648/sample-000023/original.png",
"sha256": "31770006f65cd7242a93ae2e11c5ad327db06dd1a425977073ed21c82deb3503",
"size_bytes": 1776433,
"crop_top": 0,
"... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000023/original.png",
"sha256": "31770006f65cd7242a93ae2e11c5ad327db06dd1a425977073ed21c82deb3503",
"size_bytes": 1776433,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 1,639,591,183 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 19,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000023",
"source_dataset_row_id": "anvil-17633648-sample000023-rollout01",
"source_split": "train",
"source_row_numbers": [
145,
146,
147,
... | |
anvil-17633648-sample000024 | anvil-17633648-sample000024 | anvil-17633648-sample000024 | 24 | train | cross-3 | cross-strategy (3 refs, random) | design_system_match | 3 | com.podcast.podcasts | com.podcast.podcasts__chunk_0001 | com.podcast.podcasts/com.podcast.podcasts__chunk_0001/screenshots/2024_4_26_15_14_5ceecddd8fdf4c9d97532f13a3515f03-5.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "3 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2
] | [
3,
4
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000024/context-00.png",
"sha256": "45e6e883f663d59446a9e146e0f24cd82177e03271ab2de0c7c3f51239e0a2e2",
"size_bytes": 576550,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 3,
"source_path": "assets/anvil-17633648/sample-000024/rendered.png",
"sha256": "82f89463b812ef24a9b9526e0082ee1c2e0052039606ad53d6c4d691d14d29ed",
"size_bytes": 215230,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 219,494,927 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 20,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000024",
"source_dataset_row_id": "anvil-17633648-sample000024-rollout01",
"source_split": "train",
"source_row_numbers": [
153,
154,
155,
... | |
anvil-17633648-sample000025 | anvil-17633648-sample000025 | anvil-17633648-sample000025 | 25 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.sayweee.weee | com.sayweee.weee__chunk_0000 | com.sayweee.weee/com.sayweee.weee__chunk_0000/screenshots/2024_5_6_18_13_9d79caabd27e494fa5df771587e4d60b-5.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "original",
"source_path": "assets/anvil-17633648/sample-000025/original.png",
"sha256": "633c961c6aed9d08187711ec36c9fcdeac5019e12e3325b45f03ef2a055185e6",
"size_bytes": 2352946,
"crop_top": 0,
"... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000025/original.png",
"sha256": "633c961c6aed9d08187711ec36c9fcdeac5019e12e3325b45f03ef2a055185e6",
"size_bytes": 2352946,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 815,887,704 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 21,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000025",
"source_dataset_row_id": "anvil-17633648-sample000025-rollout01",
"source_split": "train",
"source_row_numbers": [
161,
162,
163,
... | |
anvil-17633648-sample000026 | anvil-17633648-sample000026 | anvil-17633648-sample000026 | 26 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.imdb.mobile | com.imdb.mobile__chunk_0001 | com.imdb.mobile/com.imdb.mobile__chunk_0001/screenshots/2024_4_20_14_38_ec6a19f53606451f87d580d73e2fe00d-14.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000026/context-00.png",
"sha256": "aadf95d4dcf2a97f56ce1e40ec8c6c91d3c18722e9851f2207918e454a9b59ca",
"size_bytes": 389249,
"crop_top": 58,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000026/rendered.png",
"sha256": "3073b95da27266193e50e369da816951fe3012b88178b6fcb10e6122c98a7a8d",
"size_bytes": 214748,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 58 | false | [] | 770,902,370 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 22,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000026",
"source_dataset_row_id": "anvil-17633648-sample000026-rollout01",
"source_split": "train",
"source_row_numbers": [
169,
170,
171,
... | |
anvil-17633648-sample000029 | anvil-17633648-sample000029 | anvil-17633648-sample000029 | 29 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.etsy.android | com.etsy.android__chunk_0004 | com.etsy.android/com.etsy.android__chunk_0004/screenshots/2024_4_23_16_47_f3c21ac5d26641cc9dd1c642e0d697ae-21.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "rendered",
"source_path": "assets/anvil-17633648/sample-000029/rendered.png",
"sha256": "546c5ca00246dde287cc7b46643f2bcd04442c2835ae8662329c3f46311c5f06",
"size_bytes": 1623769,
"crop_top": 0,
"... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000029/rendered.png",
"sha256": "546c5ca00246dde287cc7b46643f2bcd04442c2835ae8662329c3f46311c5f06",
"size_bytes": 1623769,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 1,567,087,110 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 24,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000029",
"source_dataset_row_id": "anvil-17633648-sample000029-rollout01",
"source_split": "train",
"source_row_numbers": [
185,
186,
187,
... | |
anvil-17633648-sample000030 | anvil-17633648-sample000030 | anvil-17633648-sample000030 | 30 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.google.android.gm | com.google.android.gm__chunk_0000 | com.google.android.gm/com.google.android.gm__chunk_0000/screenshots/2024_5_6_17_9_b27819a25c1f4fb3aee3739fc59644bb-7.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000030/context-00.png",
"sha256": "97aba2329f099b015d6b8caf19eeff60209e25cafcd1a3316587179f64bcd327",
"size_bytes": 1413665,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000030/rendered.png",
"sha256": "0f08a06638a2d2cf2fff1f688a632adadcefc0d302338f9c9a97173bbe8d5d6a",
"size_bytes": 323579,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 2,089,750,213 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 25,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000030",
"source_dataset_row_id": "anvil-17633648-sample000030-rollout01",
"source_split": "train",
"source_row_numbers": [
193,
194,
195,
... | |
anvil-17633648-sample000031 | anvil-17633648-sample000031 | anvil-17633648-sample000031 | 31 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.google.android.gm | com.google.android.gm__chunk_0000 | com.google.android.gm/com.google.android.gm__chunk_0000/screenshots/2024_5_6_17_9_b27819a25c1f4fb3aee3739fc59644bb-8.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000031/context-00.png",
"sha256": "a9a3d6b071cd718799e4db4d6aad0c828e286f1b4d32c5460b6db35189a96897",
"size_bytes": 850761,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000031/rendered.png",
"sha256": "03693d2c2137afbf22b5a3c7c45441e5bef6dd0e5ae578f8fd0ae774c01914e9",
"size_bytes": 605936,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 812,896,425 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 26,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000031",
"source_dataset_row_id": "anvil-17633648-sample000031-rollout01",
"source_split": "train",
"source_row_numbers": [
201,
202,
203,
... | |
anvil-17633648-sample000033 | anvil-17633648-sample000033 | anvil-17633648-sample000033 | 33 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.cnn.mobile.android.phone | com.cnn.mobile.android.phone__chunk_0000 | com.cnn.mobile.android.phone/com.cnn.mobile.android.phone__chunk_0000/screenshots/2024_3_23_20_41_fc9bbfc483864814aa17a75a3597fd6b-9.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "rendered",
"source_path": "assets/anvil-17633648/sample-000033/rendered.png",
"sha256": "c9e7a85acba3c82ade39e8d246a792331a9d74af22542f9164709f503328c69e",
"size_bytes": 1425597,
"crop_top": 0,
"... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000033/rendered.png",
"sha256": "c9e7a85acba3c82ade39e8d246a792331a9d74af22542f9164709f503328c69e",
"size_bytes": 1425597,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 1,781,132,987 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 28,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000033",
"source_dataset_row_id": "anvil-17633648-sample000033-rollout01",
"source_split": "train",
"source_row_numbers": [
217,
218,
219,
... | |
anvil-17633648-sample000036 | anvil-17633648-sample000036 | anvil-17633648-sample000036 | 36 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | com.google.android.apps.magazines | com.google.android.apps.magazines__chunk_0000 | com.google.android.apps.magazines/com.google.android.apps.magazines__chunk_0000/screenshots/2024_4_24_21_17_89aa630439d74d5dbdf7ed0bbf1a3ad9-6.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000036/context-00.png",
"sha256": "a8cb2c0abf287fe98e72a410c0e3229bbf26840c85ef677a4915d53924fbf63b",
"size_bytes": 1520376,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000036/original.png",
"sha256": "c1c73b7c90e8a28fc3705f0637a38166a926cfb7405f6d7b0ff67ef7c584160f",
"size_bytes": 978672,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 1,239,854,340 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 30,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000036",
"source_dataset_row_id": "anvil-17633648-sample000036-rollout01",
"source_split": "train",
"source_row_numbers": [
233,
234,
235,
... | |
anvil-17633648-sample000037 | anvil-17633648-sample000037 | anvil-17633648-sample000037 | 37 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | musclebooster.workout.home.gym.abs.loseweight | musclebooster.workout.home.gym.abs.loseweight__chunk_0000 | musclebooster.workout.home.gym.abs.loseweight/musclebooster.workout.home.gym.abs.loseweight__chunk_0000/screenshots/2024_4_22_16_36_167fb3a13a144fee994407fd5f8c6bcd-6.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000037/context-00.png",
"sha256": "03ad6ba0c6d5ab0579b90f4405d624502660670010b9da46dc7ddd4dc2bd392e",
"size_bytes": 647241,
"crop_top": 68,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000037/rendered.png",
"sha256": "8f3967d0f3b5c91c90aa5519f4f920a234aed9a6cb838c3f185442878edee039",
"size_bytes": 156553,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 68 | false | [] | 1,513,056,542 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 31,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000037",
"source_dataset_row_id": "anvil-17633648-sample000037-rollout01",
"source_split": "train",
"source_row_numbers": [
241,
242,
243,
... | |
anvil-17633648-sample000038 | anvil-17633648-sample000038 | anvil-17633648-sample000038 | 38 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.zzkko | com.zzkko__chunk_0000 | com.zzkko/com.zzkko__chunk_0000/screenshots/2024_5_6_15_5_2f7ba0f9112f428ba432580961c1354c-11.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000038/context-00.png",
"sha256": "6dfb664ef1d24c795a08dd297e6b0524660d1af6d095fbe4d11ff7509990e4b8",
"size_bytes": 512139,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000038/original.png",
"sha256": "305a1be547669a14c787b7a134068c458755ea546a40a31d814ef83166d25a23",
"size_bytes": 1055477,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 98,407,155 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 32,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000038",
"source_dataset_row_id": "anvil-17633648-sample000038-rollout01",
"source_split": "train",
"source_row_numbers": [
249,
250,
251,
... | |
anvil-17633648-sample000042 | anvil-17633648-sample000042 | anvil-17633648-sample000042 | 42 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.imdb.mobile | com.imdb.mobile__chunk_0000 | com.imdb.mobile/com.imdb.mobile__chunk_0000/screenshots/2024_4_20_13_23_620f68b39c9c4945b81d35a2bae67e9a-13.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000042/context-00.png",
"sha256": "e075345618e4a4f0f4e22faf48047812c04bae1703998f1165cf5d11571deb77",
"size_bytes": 1649243,
"crop_top": 58,
"fit_... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000042/original.png",
"sha256": "d5984796db36de2dee1407b75a469ef48fc1f1b07f135c243d5d02ee10a384f4",
"size_bytes": 465137,
"crop_top": 58,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 58 | false | [] | 499,914,663 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 33,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000042",
"source_dataset_row_id": "anvil-17633648-sample000042-rollout01",
"source_split": "train",
"source_row_numbers": [
257,
258,
259,
... | |
anvil-17633648-sample000043 | anvil-17633648-sample000043 | anvil-17633648-sample000043 | 43 | train | cross-7 | cross-strategy (7 refs, random) | design_system_match | 7 | bbc.mobile.sport.ww | bbc.mobile.sport.ww__chunk_0001 | bbc.mobile.sport.ww/bbc.mobile.sport.ww__chunk_0001/screenshots/2024_4_25_19_58_c62044e159594dc6b2cdc9c2d4f92e3d-9.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "7 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4,
5,
6
] | [
7,
8
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000043/context-00.png",
"sha256": "a545b430cc4416be3377eccd65ba70cfc9e1bcf36ee840713fcd4251316c5e4c",
"size_bytes": 995685,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 7,
"source_path": "assets/anvil-17633648/sample-000043/original.png",
"sha256": "6298ad7e68185caacbeb9b9ca1768c6b383c258f74e2ec7d2bb0b40b370a02d5",
"size_bytes": 386207,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 596,943,816 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 34,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000043",
"source_dataset_row_id": "anvil-17633648-sample000043-rollout01",
"source_split": "train",
"source_row_numbers": [
265,
266,
267,
... | |
anvil-17633648-sample000045 | anvil-17633648-sample000045 | anvil-17633648-sample000045 | 45 | train | cross-5 | cross-strategy (5 refs, random) | design_system_match | 5 | com.microsoft.teams | com.microsoft.teams__chunk_0000 | com.microsoft.teams/com.microsoft.teams__chunk_0000/screenshots/2024_4_24_21_18_f284f97becb6477d9aa9731295908d18-3.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4
] | [
5,
6
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000045/context-00.png",
"sha256": "06fb623da858b462886769a32369165d88f346670088a899a1c9947b24d1ac0f",
"size_bytes": 595442,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 5,
"source_path": "assets/anvil-17633648/sample-000045/original.png",
"sha256": "9ff9032a34d7025a6ab159d9ad9ebace4ea18de586acf2374e1bf60d78e14b03",
"size_bytes": 470076,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 783,471,182 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 36,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000045",
"source_dataset_row_id": "anvil-17633648-sample000045-rollout01",
"source_split": "train",
"source_row_numbers": [
281,
282,
283,
... | |
anvil-17633648-sample000047 | anvil-17633648-sample000047 | anvil-17633648-sample000047 | 47 | train | cross-5 | cross-strategy (5 refs, random) | design_system_match | 5 | com.yelp.android | com.yelp.android__chunk_0000 | com.yelp.android/com.yelp.android__chunk_0000/screenshots/2024_6_19_18_17_9cecfe11a8f9473b9169e04bbf7e1a27-12.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or ... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "5 reference screenshots will be attached first, followed by two candidates. The reference image(s) come from the same design system as the... | [
0,
1,
2,
3,
4
] | [
5,
6
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000047/context-00.png",
"sha256": "1b187b0be1204da922d7b811d6190eafa6c79b07dbf72366b05df5abdfcd9c1e",
"size_bytes": 1376258,
"crop_top": 0,
"fit_s... | [
{
"position": 0,
"label": "A",
"kind": "original",
"prompt_index": 5,
"source_path": "assets/anvil-17633648/sample-000047/original.png",
"sha256": "68897c2fda6e595079f2c7e77a1930fb6a1aa015a0b53ccbeeab075b96932b34",
"size_bytes": 788558,
"crop_top": 0,
"fit_size": []
},
{
... | 0 | A | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 449,912,967 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 37,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000047",
"source_dataset_row_id": "anvil-17633648-sample000047-rollout01",
"source_split": "train",
"source_row_numbers": [
289,
290,
291,
... | |
anvil-17633648-sample000048 | anvil-17633648-sample000048 | anvil-17633648-sample000048 | 48 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.booking | com.booking__chunk_0000 | com.booking/com.booking__chunk_0000/screenshots/2024_5_1_16_6_480de1c88f3f49df9537cbbdcdb680f1-9.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000048/context-00.png",
"sha256": "05bd64e27692296ac96aac0cbe4acc952d809aafdfc179dab82250aa68dac234",
"size_bytes": 291129,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000048/rendered.png",
"sha256": "a92f401ea35db689f89891b217c22a7171c724538eb1633a40bf36bd1cdf04f0",
"size_bytes": 211928,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 573,330,547 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 38,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000048",
"source_dataset_row_id": "anvil-17633648-sample000048-rollout01",
"source_split": "train",
"source_row_numbers": [
297,
298,
299,
... | |
anvil-17633648-sample000049 | anvil-17633648-sample000049 | anvil-17633648-sample000049 | 49 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.podcast.podcasts | com.podcast.podcasts__chunk_0001 | com.podcast.podcasts/com.podcast.podcasts__chunk_0001/screenshots/2024_4_26_10_17_25b2034df9c54923b8af0490b742f144-9.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000049/context-00.png",
"sha256": "3fb4cb66d0cd50cdf84ddb952c4c40ce167f33715950ce82a0825794b74665fb",
"size_bytes": 1352228,
"crop_top": 68,
"fit_... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000049/rendered.png",
"sha256": "be42e9a909fd2e32f9bbb8a4b4780d7f97e55aaae067304bb8729d15a9003327",
"size_bytes": 1192188,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 68 | false | [] | 2,011,450,454 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 39,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000049",
"source_dataset_row_id": "anvil-17633648-sample000049-rollout01",
"source_split": "train",
"source_row_numbers": [
305,
306,
307,
... | |
anvil-17633648-sample000050 | anvil-17633648-sample000050 | anvil-17633648-sample000050 | 50 | train | cross-0 | cross-strategy (0 refs, random) | real_vs_generated | 0 | com.spotify.music | com.spotify.music__chunk_0000 | com.spotify.music/com.spotify.music__chunk_0000/screenshots/2024_4_9_2_10_c32f60d80721433bbe4fb43cb450f2b8-14.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the HTML-rendered output. First, reason step by step about which candidate is the original. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. reason should explain s... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "Two images will be attached. Image 1 is Candidate A and image 2 is Candidate B. One is the original source screenshot and the other is the... | [] | [
0,
1
] | [
{
"prompt_index": 0,
"role": "candidate",
"candidate_label": "A",
"candidate_kind": "rendered",
"source_path": "assets/anvil-17633648/sample-000050/rendered.png",
"sha256": "009bc502dbbc4a530f06f726ed5631e82c30ba7a526e56224283b79ae423dbc1",
"size_bytes": 793167,
"crop_top": 0,
"f... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 0,
"source_path": "assets/anvil-17633648/sample-000050/rendered.png",
"sha256": "009bc502dbbc4a530f06f726ed5631e82c30ba7a526e56224283b79ae423dbc1",
"size_bytes": 793167,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | null | false | [] | 1,391,645,448 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 40,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000050",
"source_dataset_row_id": "anvil-17633648-sample000050-rollout01",
"source_split": "train",
"source_row_numbers": [
313,
314,
315,
... | |
anvil-17633648-sample000051 | anvil-17633648-sample000051 | anvil-17633648-sample000051 | 51 | train | cross-1 | cross-strategy (1 refs, random) | design_system_match | 1 | com.google.android.apps.maps | com.google.android.apps.maps__chunk_0000 | com.google.android.apps.maps/com.google.android.apps.maps__chunk_0000/screenshots/2024_5_6_20_43_e07efd79f09e4aacbec9032893c2b319-11.png | You are a strict visual judge. Follow the user's instructions exactly and return JSON only. | 1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the target. First, reason step by step about which candidate better follows the design system. Then output JSON only in this shape: {"reason":"...","choice":"A|B"}. choice must be A or B... | [
{
"role": "system",
"content": "You are a strict visual judge. Follow the user's instructions exactly and return JSON only."
},
{
"role": "user",
"content": "1 reference screenshot will be attached first, followed by two candidates. The reference image(s) come from the same design system as the ... | [
0
] | [
1,
2
] | [
{
"prompt_index": 0,
"role": "context",
"candidate_label": null,
"candidate_kind": null,
"source_path": "assets/anvil-17633648/sample-000051/context-00.png",
"sha256": "a7ed14514511f2857d4fd811684aa28239092d556b28cf7278cf7bf954a71d6b",
"size_bytes": 439106,
"crop_top": 0,
"fit_si... | [
{
"position": 0,
"label": "A",
"kind": "rendered",
"prompt_index": 1,
"source_path": "assets/anvil-17633648/sample-000051/rendered.png",
"sha256": "7fdec8f42aa6999e1d32c9c4cb867e8de4090496856b7a21f3ae02f6c6cb987b",
"size_bytes": 150559,
"crop_top": 0,
"fit_size": [
1440,
... | 1 | B | original | gemma4:26b-a4b-it-q4_K_M | http://127.0.0.1:23648/api/chat | 1 | 0.95 | 64 | 4,096 | null | 0 | false | [] | 1,308,098,925 | null | null | {
"source_jsonl": "train.jsonl",
"source_row_number": 41,
"source_root": "judge_prompt_dataset_v1_dedup",
"source_sample_row_id": "anvil-17633648-sample000051",
"source_dataset_row_id": "anvil-17633648-sample000051-rollout01",
"source_split": "train",
"source_row_numbers": [
321,
322,
323,
... |
End of preview. Expand in Data Studio
No dataset card yet
- Downloads last month
- 41