JJimmyL commited on
Commit
9fab2fd
·
verified ·
1 Parent(s): 9ac7568

Publish organized silent demonstrations and task inputs 27/30

Browse files
tasks/E39/C/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E39C 文字规则
2
+
3
+ 下排颜色次序和各T朝向都与上排一致
tasks/E40/A/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E40A · 在指定图形处夹长尾夹
2
+
3
+ ## 执行任务目标
4
+
5
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E40/A/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:226a7ccfa8e349ad3cc5c6da58d4624453725ca699ff81096cafd6c2cf235c93
3
+ size 5304829
tasks/E40/A/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:352929db8c0ec98a974a4f90c2fbf60939623c0d494bf395a5257d6051b1d6b7
3
+ size 1442230
tasks/E40/A/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E40",
3
+ "variant": "A",
4
+ "local_rule": "夹住圆形和叉号对应的盘边",
5
+ "original_task_title": "在指定图形处夹长尾夹",
6
+ "original_inference_prompt": "按照示范的图形选择,在对应盘边夹上长尾夹。",
7
+ "original_evaluation_scope": "正确图形边缘被真实夹口夹住并留夹,未选边无夹。",
8
+ "original_scope_note": "虽然题单F09有删除标记,实录仍存在。真实道具为黑色长尾夹,夹在带图形标签的托盘外沿,夹臂需张开再释放;不是颜色衣夹。",
9
+ "reviewed_rule_evidence": "2.70s圆,7.53s叉号",
10
+ "execution_design_v3": {
11
+ "title": "在指定图形处夹长尾夹",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按照示范的图形选择,在对应盘边夹上长尾夹。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E40/A/real.mp4",
42
+ "sha256": "226a7ccfa8e349ad3cc5c6da58d4624453725ca699ff81096cafd6c2cf235c93",
43
+ "bytes": 5304829,
44
+ "duration_seconds": 8.066667,
45
+ "frame_count": 242,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "f97103299e709ef63a4bd28624f1dead18344d1f40dab3f2dcba334f41d11b71",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E40/E40_A__IMG_0625.MOV",
56
+ "sha256": "05902a3cf82858d8f255da8e01ac7c2df49ffe4933e677e86549b7e3149af4ed"
57
+ },
58
+ "source_video_stream_sha256": "353546e1ad5d9e83f641f4c6f599708808d7f54e0d1773eaf73c2d3fb724bd7d",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E40/A/sim.mp4",
70
+ "sha256": "352929db8c0ec98a974a4f90c2fbf60939623c0d494bf395a5257d6051b1d6b7",
71
+ "bytes": 1442230,
72
+ "duration_seconds": 69.28,
73
+ "frame_count": 1732,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "6c4ac5a1cd11def6a9f4cf34263723108afeee2691a9dc792539aa7f59fd681a",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E40/E40_A__head.mp4",
84
+ "sha256": "4fee6fc48491740acded8fc82645d7e9a9715f4cfa12b42d35ac6bc36a3c3d73"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E40/A/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E40A 文字规则
2
+
3
+ 夹住圆形和叉号对应的盘边
tasks/E40/B/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E40B · 在指定图形处夹长尾夹
2
+
3
+ ## 执行任务目标
4
+
5
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E40/B/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:325670b300979aa1242319906e0f6f5bf9548473fe07492699c9e6c7a681e2b1
3
+ size 5192814
tasks/E40/B/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a58cf87cad463118ce10db8f6e5ef87d3a860d6f43d1df2950badf282435ee6
3
+ size 1452254
tasks/E40/B/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E40",
3
+ "variant": "B",
4
+ "local_rule": "夹住方形和三角形对应的盘边",
5
+ "original_task_title": "在指定图形处夹长尾夹",
6
+ "original_inference_prompt": "按照示范的图形选择,在对应盘边夹上长尾夹。",
7
+ "original_evaluation_scope": "正确图形边缘被真实夹口夹住并留夹,未选边无夹。",
8
+ "original_scope_note": "虽然题单F09有删除标记,实录仍存在。真实道具为黑色长尾夹,夹在带图形标签的托盘外沿,夹臂需张开再释放;不是颜色衣夹。",
9
+ "reviewed_rule_evidence": "3.20s方,7.43s三角",
10
+ "execution_design_v3": {
11
+ "title": "在指定图形处夹长尾夹",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按照示范的图形选择,在对应盘边夹上长尾夹。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E40/B/real.mp4",
42
+ "sha256": "325670b300979aa1242319906e0f6f5bf9548473fe07492699c9e6c7a681e2b1",
43
+ "bytes": 5192814,
44
+ "duration_seconds": 7.933333,
45
+ "frame_count": 238,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "a321505a7c731416955b41a95206527d3e8f7b95435488b2299916acd72cff08",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E40/E40_B__IMG_0626.MOV",
56
+ "sha256": "6f63bbb7ebf63f041252c9c5c9c89393f729e85fe115ae02200721e9f8110d37"
57
+ },
58
+ "source_video_stream_sha256": "e6121723cfc1de112dc25308430f41820d96373661f3e8f7116863c8ada1c440",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E40/B/sim.mp4",
70
+ "sha256": "3a58cf87cad463118ce10db8f6e5ef87d3a860d6f43d1df2950badf282435ee6",
71
+ "bytes": 1452254,
72
+ "duration_seconds": 69.2,
73
+ "frame_count": 1730,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "2a0a683346a0e0813ce9a9fed0f25f3f1942f543f0dd9052ec53cdac8d6c2227",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E40/E40_B__head.mp4",
84
+ "sha256": "52ce19175bec7831eff6d7b9ce1e684f402e55ec563d211a25a0b81b55a3652e"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E40/B/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E40B 文字规则
2
+
3
+ 夹住方形和三角形对应的盘边
tasks/E40/C/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E40C · 在指定图形处夹长尾夹
2
+
3
+ ## 执行任务目标
4
+
5
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按照示范的图形选择,在对应盘边夹上长尾夹。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E40/C/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2850273f60b97bf89675a9943219192bf06a50799f0e0fb8f71b1a4fea825701
3
+ size 6929067
tasks/E40/C/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52a045114cab83579a780b4139d2dd14a18e503a257ca4d9a4a77924cbb21137
3
+ size 2102144
tasks/E40/C/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E40",
3
+ "variant": "C",
4
+ "local_rule": "夹住圆形、三角形和叉号对应的盘边",
5
+ "original_task_title": "在指定图形处夹长尾夹",
6
+ "original_inference_prompt": "按照示范的图形选择,在对应盘边夹上长尾夹。",
7
+ "original_evaluation_scope": "正确图形边缘被真实夹口夹住并留夹,未选边无夹。",
8
+ "original_scope_note": "虽然题单F09有删除标记,实录仍存在。真实道具为黑色长尾夹,夹在带图形标签的托盘外沿,夹臂需张开再释放;不是颜色衣夹。",
9
+ "reviewed_rule_evidence": "2.67s圆,4.67s三角,8.00s叉号",
10
+ "execution_design_v3": {
11
+ "title": "在指定图形处夹长尾夹",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按照示范的图形选择,在对应盘边夹上长尾夹。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E40/C/real.mp4",
42
+ "sha256": "2850273f60b97bf89675a9943219192bf06a50799f0e0fb8f71b1a4fea825701",
43
+ "bytes": 6929067,
44
+ "duration_seconds": 10.0,
45
+ "frame_count": 300,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "cdd4855df3d2f2ee05347b293ca35a3096d2ac410e0a02e144464c1f38740a4d",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E40/E40_C__IMG_0627.MOV",
56
+ "sha256": "25ec46faa4784cfd923c0ad275a3e2b64230f427d970d56959e01a964e9734e6"
57
+ },
58
+ "source_video_stream_sha256": "5354f27f23ac5d6c391dee7ed160dae1fa2a9ced5eca9b574286676af5fce795",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E40/C/sim.mp4",
70
+ "sha256": "52a045114cab83579a780b4139d2dd14a18e503a257ca4d9a4a77924cbb21137",
71
+ "bytes": 2102144,
72
+ "duration_seconds": 100.28,
73
+ "frame_count": 2507,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "202cf43e9dd31184e4313ce2caa6253c9a40961c4702a1f5515ee7073c57b1c9",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E40/E40_C__head.mp4",
84
+ "sha256": "058d195e31c6388f73ac8688005464f4db4069960e2ac567a73b8709d855deeb"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E40/C/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E40C 文字规则
2
+
3
+ 夹住圆形、三角形和叉号对应的盘边
tasks/E41/A/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E41A · 为不同长度木条接续一段
2
+
3
+ ## 执行任务目标
4
+
5
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E41/A/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64d24b45389faf58b833c331109692c93eeae4017d6eca08b0732313ddf127de
3
+ size 6090146
tasks/E41/A/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e2010df13b0ecb6cc197ec60465e436319c8e289c8a5f21965e9dcd9b343cd8
3
+ size 1680912
tasks/E41/A/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E41",
3
+ "variant": "A",
4
+ "local_rule": "短前段接长备件,中前段接中备件,长前段接短备件",
5
+ "original_task_title": "为不同长度木条接续一段",
6
+ "original_inference_prompt": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
7
+ "original_evaluation_scope": "短中长映射正确、端点邻接且方向共线;不要求总长精确相同。",
8
+ "original_scope_note": "参照行初始从左到右短中长。直接记前段长度与备件长度对应,不用题单手写升降序等含混描述;不要求总长严格相等,视频尺寸并未精确量齐。",
9
+ "reviewed_rule_evidence": "2.70/5.40/8.10s备件从左到右长中短",
10
+ "execution_design_v3": {
11
+ "title": "为不同长度木条接续一段",
12
+ "difficulty": "medium",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
20
+ "input_change_description": "三组全新的长度值,保持短中长对应",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E41/A/real.mp4",
42
+ "sha256": "64d24b45389faf58b833c331109692c93eeae4017d6eca08b0732313ddf127de",
43
+ "bytes": 6090146,
44
+ "duration_seconds": 10.1,
45
+ "frame_count": 303,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "04180aeec4275f992be2fe70e8a3b2f22ffa1e1ce9b83b1d6336e0322a023028",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E41/E41_A__IMG_0628.MOV",
56
+ "sha256": "fda45ed3c8037621588e7ab09e13af0e69e6d91c497392dfdd89b22047c8cc7c"
57
+ },
58
+ "source_video_stream_sha256": "d78ba664f7522ec8cbe2aeefd9208b3ba24562b992ace9d52b69291e6f682f9c",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E41/A/sim.mp4",
70
+ "sha256": "8e2010df13b0ecb6cc197ec60465e436319c8e289c8a5f21965e9dcd9b343cd8",
71
+ "bytes": 1680912,
72
+ "duration_seconds": 58.4,
73
+ "frame_count": 1460,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "3111709d416eb0d1fe913686b8ef501fd437c29b057341cf081fd033b1b9fc59",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E41/E41_A__head.mp4",
84
+ "sha256": "e42cc9f4e53279c1341bd7cdf3070c261a4a10ab261b66521f8b0a2103f4df71"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E41/A/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E41A 文字规则
2
+
3
+ 短前段接长备件,中前段接中备件,长前段接短备件
tasks/E41/B/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E41B · 为不同长度木条接续一段
2
+
3
+ ## 执行任务目标
4
+
5
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E41/B/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bef212a674a9d1dae244d2fe30511742fe699419457ce97a730ed44600e10af5
3
+ size 5088541
tasks/E41/B/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26d6867d3079029b191a7ea8c6604dae0f6a848b50d88c6037175b2482660dcc
3
+ size 1758368
tasks/E41/B/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E41",
3
+ "variant": "B",
4
+ "local_rule": "短接短、中接中、长接长",
5
+ "original_task_title": "为不同长度木条接续一段",
6
+ "original_inference_prompt": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
7
+ "original_evaluation_scope": "短中长映射正确、端点邻接且方向共线;不要求总长精确相同。",
8
+ "original_scope_note": "参照行初始从左到右短中长。直接记前段长度与备件长度对应,不用题单手写升降序等含混描述;不要求总长严格相等,视频尺寸并未精确量齐。",
9
+ "reviewed_rule_evidence": "3.53/5.27/7.60s备件从左到右短中长",
10
+ "execution_design_v3": {
11
+ "title": "为不同长度木条接续一段",
12
+ "difficulty": "medium",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
20
+ "input_change_description": "三组全新的长度值,保持短中长对应",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E41/B/real.mp4",
42
+ "sha256": "bef212a674a9d1dae244d2fe30511742fe699419457ce97a730ed44600e10af5",
43
+ "bytes": 5088541,
44
+ "duration_seconds": 8.766667,
45
+ "frame_count": 263,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "f2b5f95c3637a9f8547e18d7ba4fdeeeb4526d36a5b0e223a3341439fbce63a4",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E41/E41_B__IMG_0629.MOV",
56
+ "sha256": "3575e3f9c89ddbac4e924805c809f0fc0695c6cba05fc3fac41d0bfe7180eb42"
57
+ },
58
+ "source_video_stream_sha256": "0b5e582228cd4c151e6cac1b237e2a37176dfab3e15a3e5668c673d6b62a0b46",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E41/B/sim.mp4",
70
+ "sha256": "26d6867d3079029b191a7ea8c6604dae0f6a848b50d88c6037175b2482660dcc",
71
+ "bytes": 1758368,
72
+ "duration_seconds": 60.96,
73
+ "frame_count": 1524,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "23ce41c50e18caf07f5ec2f619ba01123c13046f70d039405ea5f1b0340b717c",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E41/E41_B__head.mp4",
84
+ "sha256": "1f9e4fc2bbf0839a4939930674a840492c9b3f10ab5fb464bd04a3c032c06f95"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E41/B/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E41B 文字规则
2
+
3
+ 短接短、中接中、长接长
tasks/E41/C/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E41C · 为不同长度木条接续一段
2
+
3
+ ## 执行任务目标
4
+
5
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E41/C/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8efca5d9cd74ee5f298872d21b14e367c3f5d34ae163871025b74892d2b35ca
3
+ size 5481497
tasks/E41/C/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67cd00510e4e650e32f536911e440702044c8ccd16a63b4d8c7bf406eee91068
3
+ size 1636192
tasks/E41/C/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E41",
3
+ "variant": "C",
4
+ "local_rule": "短接长、中接短、长接中",
5
+ "original_task_title": "为不同长度木条接续一段",
6
+ "original_inference_prompt": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
7
+ "original_evaluation_scope": "短中长映射正确、端点邻接且方向共线;不要求总长精确相同。",
8
+ "original_scope_note": "参照行初始从左到右短中长。直接记前段长度与备件长度对应,不用题单手写升降序等含混描述;不要求总长严格相等,视频尺寸并未精确量齐。",
9
+ "reviewed_rule_evidence": "2.53/4.43/7.60s备件从左到右长短中",
10
+ "execution_design_v3": {
11
+ "title": "为不同长度木条接续一段",
12
+ "difficulty": "medium",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按示范中前段长度与接续段长度的对应关系,为每根木条接上一段。",
20
+ "input_change_description": "三组全新的长度值,保持短中长对应",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E41/C/real.mp4",
42
+ "sha256": "e8efca5d9cd74ee5f298872d21b14e367c3f5d34ae163871025b74892d2b35ca",
43
+ "bytes": 5481497,
44
+ "duration_seconds": 9.5,
45
+ "frame_count": 285,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "716628a7ff57741ce7e9a140397bea92576d88033e7599cca0e387d86b73085e",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E41/E41_C__IMG_0630.MOV",
56
+ "sha256": "13634fc5b8a277a49aead1b91afe0acdeabcb14e1cc0dba4673c264f1fea643a"
57
+ },
58
+ "source_video_stream_sha256": "d9665d4e1aa950cb6490d92dd3eab35f1fb2ed793700c76959e1845d4f3c5bd3",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E41/C/sim.mp4",
70
+ "sha256": "67cd00510e4e650e32f536911e440702044c8ccd16a63b4d8c7bf406eee91068",
71
+ "bytes": 1636192,
72
+ "duration_seconds": 56.48,
73
+ "frame_count": 1412,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "d42915e7e32567677b699d06d419fd1732f4a84a1981223bb7f775db88fe50ad",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E41/E41_C__head.mp4",
84
+ "sha256": "f49c00d87a3aa2bc8334be299c1a4601421ec8514ce8e7b7e7bc854041b467a8"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E41/C/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E41C 文字规则
2
+
3
+ 短接长、中接短、长接中
tasks/E42/A/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E42A · 用铲沿推球到图形区域
2
+
3
+ ## 执行任务目标
4
+
5
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E42/A/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5db6811df6f37c19a0801e7825237db9cdd7a1d4c44725df384fd869fc08b229
3
+ size 6822747
tasks/E42/A/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96f151ce32a7299ced92bc05a7d64844de9fb2f02a79241222b9b9da5a4a0c30
3
+ size 1831481
tasks/E42/A/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E42",
3
+ "variant": "A",
4
+ "local_rule": "把两球推到圆标记纸区域",
5
+ "original_task_title": "用铲沿推球到图形区域",
6
+ "original_inference_prompt": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
7
+ "original_evaluation_scope": "两球在目标纸区域并停稳,过程有铲沿推动;允许纸边容差。",
8
+ "original_scope_note": "真实工具是平铲不是刷子,使用侧沿推球而不是托起搬运。C末态在纸边,不完全位于画出的方框内;以标记纸区域含合理边界容差为目标,不能标成严格图形内部成功。",
9
+ "reviewed_rule_evidence": "9.43s两球停在圆纸上",
10
+ "execution_design_v3": {
11
+ "title": "用铲沿推球到图形区域",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E42/A/real.mp4",
42
+ "sha256": "5db6811df6f37c19a0801e7825237db9cdd7a1d4c44725df384fd869fc08b229",
43
+ "bytes": 6822747,
44
+ "duration_seconds": 10.1,
45
+ "frame_count": 303,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "436fd4804a46d626d47f87bd4d7985ab275ca83d8bde90dee7e0f97fe29e8e16",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E42/E42_A__IMG_0631.MOV",
56
+ "sha256": "9a381f8375f4dad97c82e7a5068d3c5a703884c017975324bed80c788fc2c24c"
57
+ },
58
+ "source_video_stream_sha256": "ed4d02e33af89af7cc13fa3afca261935644c2ea629889bfc91d16d6f61dea37",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E42/A/sim.mp4",
70
+ "sha256": "96f151ce32a7299ced92bc05a7d64844de9fb2f02a79241222b9b9da5a4a0c30",
71
+ "bytes": 1831481,
72
+ "duration_seconds": 94.8,
73
+ "frame_count": 2370,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "97d13703fbd66a7f027855bc69d60c296b9aa86ab66a048d90e03a4947a89211",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E42/E42_A__head.mp4",
84
+ "sha256": "3286450a67becdb1783a7ca3016ca789f7d6ed15c8d1a17bd86390ad61d274a7"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E42/A/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E42A 文字规则
2
+
3
+ 把两球推到圆标记纸区域
tasks/E42/B/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E42B · 用铲沿推球到图形区域
2
+
3
+ ## 执行任务目标
4
+
5
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E42/B/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:427088918ca6c78b40d4277ba00848d1a5ad035f4137860e3def5509a26de025
3
+ size 4213038
tasks/E42/B/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3288c669664b403b5d3f8dd90da9d4f36d2435465f65dc3e322fcbde351c57a0
3
+ size 1526148
tasks/E42/B/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E42",
3
+ "variant": "B",
4
+ "local_rule": "把两球推到三角标记纸区域",
5
+ "original_task_title": "用铲沿推球到图形区域",
6
+ "original_inference_prompt": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
7
+ "original_evaluation_scope": "两球在目标纸区域并停稳,过程有铲沿推动;允许纸边容差。",
8
+ "original_scope_note": "真实工具是平铲不是刷子,使用侧沿推球而不是托起搬运。C末态在纸边,不完全位于画出的方框内;以标记纸区域含合理边界容差为目标,不能标成严格图形内部成功。",
9
+ "reviewed_rule_evidence": "6.60s两球停在三角纸上",
10
+ "execution_design_v3": {
11
+ "title": "用铲沿推球到图形区域",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E42/B/real.mp4",
42
+ "sha256": "427088918ca6c78b40d4277ba00848d1a5ad035f4137860e3def5509a26de025",
43
+ "bytes": 4213038,
44
+ "duration_seconds": 7.6,
45
+ "frame_count": 228,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "4932bdf8343854e7706d9ac162297cee1461a26e8f7f7a99d51609f0912c5d85",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E42/E42_B__IMG_0632.MOV",
56
+ "sha256": "e7715f89b2129f644acd7d692ca3b35d7fccccc66ba7afbfd7fb34bbf5726d62"
57
+ },
58
+ "source_video_stream_sha256": "721981e0c58c7b8903ce0d4dab36d9ebac104e0da2f5b806c18da91bffe117c4",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E42/B/sim.mp4",
70
+ "sha256": "3288c669664b403b5d3f8dd90da9d4f36d2435465f65dc3e322fcbde351c57a0",
71
+ "bytes": 1526148,
72
+ "duration_seconds": 79.4,
73
+ "frame_count": 1985,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "ec5e5a07f37333ce8bd6d383e6f59cc9e13a00bbf2b20f613508b048a46f1014",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E42/E42_B__head.mp4",
84
+ "sha256": "ee69867d001d8f34fc8e7076f43f5e668bb9e6c0e5c44aba029697cf28d024cf"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E42/B/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E42B 文字规则
2
+
3
+ 把两球推到三角标记纸区域
tasks/E42/C/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E42C · 用铲沿推球到图形区域
2
+
3
+ ## 执行任务目标
4
+
5
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E42/C/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac1ab03454b426886d18aee71b01b5f0065a9017ff442e4f3b70c02b85664482
3
+ size 7907761
tasks/E42/C/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16bd9ebad8aad156990dcd76f04257b02bc37ce18c354b4afc7a0b4229e0052b
3
+ size 1685487
tasks/E42/C/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E42",
3
+ "variant": "C",
4
+ "local_rule": "把两球推到方标记纸区域",
5
+ "original_task_title": "用铲沿推球到图形区域",
6
+ "original_inference_prompt": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
7
+ "original_evaluation_scope": "两球在目标纸区域并停稳,过程有铲沿推动;允许纸边容差。",
8
+ "original_scope_note": "真实工具是平铲不是刷子,使用侧沿推球而不是托起搬运。C末态在纸边,不完全位于画出的方框内;以标记纸区域含合理边界容差为目标,不能标成严格图形内部成功。",
9
+ "reviewed_rule_evidence": "11.70–12.50s两球停在方纸边缘",
10
+ "execution_design_v3": {
11
+ "title": "用铲沿推球到图形区域",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "用平铲的侧沿,把两个球推到示范选择的图形标记纸区域。",
20
+ "input_change_description": "新对象实例和初始排列;对应关系、顺序或数量规则不变",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E42/C/real.mp4",
42
+ "sha256": "ac1ab03454b426886d18aee71b01b5f0065a9017ff442e4f3b70c02b85664482",
43
+ "bytes": 7907761,
44
+ "duration_seconds": 12.533333,
45
+ "frame_count": 376,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "44a4bba45458e5e55cf066fc2db0383670a2fabcdf8887578b54a31f6d54b0b1",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E42/E42_C__IMG_0634.MOV",
56
+ "sha256": "8c765803c614edf8a98c5c4ac8a01b80358b8fc588a93749916aaa60e0f03c65"
57
+ },
58
+ "source_video_stream_sha256": "c43da515586a43a4f45bb2aaca51b400b005cc7ca6a311924f61e7adfa79f27d",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E42/C/sim.mp4",
70
+ "sha256": "16bd9ebad8aad156990dcd76f04257b02bc37ce18c354b4afc7a0b4229e0052b",
71
+ "bytes": 1685487,
72
+ "duration_seconds": 102.48,
73
+ "frame_count": 2562,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "8a1f202e32ebee845946c29247c749fa660363c330bc4ee31470dee309a61494",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E42/E42_C__head.mp4",
84
+ "sha256": "2ef34d30055b794ef6fe1397f14cb34ae37a1c8ea30a25f116df5fd09456094f"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }
tasks/E42/C/text_rule.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ # E42C 文字规则
2
+
3
+ 把两球推到方标记纸区域
tasks/E43/A/prompt.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # E43A · 用导向条让球经过图形区域
2
+
3
+ ## 执行任务目标
4
+
5
+ 按示范选择的图形区域布置导向条,使释放后的球经过该区域。
6
+
7
+ ## 原始局部示范任务
8
+
9
+ 按示范选择的图形区域布置导向条,使释放后的球经过该区域。
10
+
11
+ 逐回合实际输入以 benchmark/assignments.jsonl 中的 instruction、text_rule 和视频 SHA-256 为准。
tasks/E43/A/real.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:743ca217fc9de5d4f46ff14b4b71f52a5d0c7beaafb7555563976f7ff7a3c1f4
3
+ size 7450124
tasks/E43/A/sim.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6adcbcc4df387a312c4d22ed4da117dbd1eae30c91f712ca38a516743602e443
3
+ size 666359
tasks/E43/A/task.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "E43",
3
+ "variant": "A",
4
+ "local_rule": "导向圆标记区域",
5
+ "original_task_title": "用导向条让球经过图形区域",
6
+ "original_inference_prompt": "按示范选择的图形区域布置导向条,使释放后的球经过该区域。",
7
+ "original_evaluation_scope": "释放后真实轨迹穿越目标区域;不要求停留;不得手抓球交付。",
8
+ "original_scope_note": "6 fps复查确认轨迹经过目标纸区域,不是停留;B跨区域的瞬间位于采样帧之间,保留5.67–5.83s区间。正式判定应检测穿越目标区域事件,不能要求最终停稳。另一人负责释放/倾斜起始板;仿真需独立固定的放球机构。导向条必须通过碰撞改变轨迹,不能按答案直接送球。",
9
+ "reviewed_rule_evidence": "7.33s释放;8.50s经过圆纸,8.67s滚出",
10
+ "execution_design_v3": {
11
+ "title": "用导向条让球经过图形区域",
12
+ "difficulty": "hard",
13
+ "supported_environments": [
14
+ "R",
15
+ "I",
16
+ "E"
17
+ ],
18
+ "main_environment": "I",
19
+ "public_goal": "按示范选择的图形区域布置导向条,使释放后的球经过该区域。",
20
+ "input_change_description": "新球起点和目标标记布局;导向条须重新布置",
21
+ "environment_base": "I",
22
+ "environment_pairing": {
23
+ "environment_base": "I",
24
+ "paired_input_required": true,
25
+ "invariants": [
26
+ "private_rule",
27
+ "execution_goal",
28
+ "object_semantics",
29
+ "rule_bearing_colors_shapes",
30
+ "task_coordinate_references"
31
+ ],
32
+ "allowed_changes": [
33
+ "workspace_appearance",
34
+ "background",
35
+ "lighting",
36
+ "rule_preserving_spatial_layout"
37
+ ]
38
+ }
39
+ },
40
+ "real": {
41
+ "path": "tasks/E43/A/real.mp4",
42
+ "sha256": "743ca217fc9de5d4f46ff14b4b71f52a5d0c7beaafb7555563976f7ff7a3c1f4",
43
+ "bytes": 7450124,
44
+ "duration_seconds": 10.033333,
45
+ "frame_count": 301,
46
+ "fps": "30/1",
47
+ "width": 1920,
48
+ "height": 1080,
49
+ "audio_tracks": 0,
50
+ "video_codec": "h264",
51
+ "video_stream_sha256": "9d7b2597dad47c740c7eaca069da259c41acbd3a115f50a00278cf587ebd338e",
52
+ "source": {
53
+ "repo": "https://www.modelscope.ai/datasets/li-jm25/VideoICL_Embodied_Set.git",
54
+ "revision": "91a4f75441065c01d94da57fc95233f2000dd1cb",
55
+ "path": "media/E43/E43_A__IMG_0636.mov",
56
+ "sha256": "1f57d5b5eb9a5bd828c87cd62b8fe94894b04b9047a0dc94616dc81baa486831"
57
+ },
58
+ "source_video_stream_sha256": "3a20e7acc971e101fa7832ff6755782cbd27f85e995f35fb48da4bd2af480ead",
59
+ "video_stream_preserved": false,
60
+ "transformation": "HEVC_to_H264_VideoToolbox_6Mbps_audio_removed",
61
+ "validation": [
62
+ "frame_count_matches_source",
63
+ "duration_difference_at_most_0.12_seconds",
64
+ "display_orientation_preserved",
65
+ "zero_audio_tracks"
66
+ ]
67
+ },
68
+ "sim": {
69
+ "path": "tasks/E43/A/sim.mp4",
70
+ "sha256": "6adcbcc4df387a312c4d22ed4da117dbd1eae30c91f712ca38a516743602e443",
71
+ "bytes": 666359,
72
+ "duration_seconds": 29.32,
73
+ "frame_count": 733,
74
+ "fps": "25/1",
75
+ "width": 640,
76
+ "height": 480,
77
+ "audio_tracks": 0,
78
+ "video_codec": "h264",
79
+ "video_stream_sha256": "77cc5ea0fa0a8fca950ee473f763b95e2e84b300aebb373f6c3d2b7248f0cfe8",
80
+ "source": {
81
+ "repo": "JJimmyL/VideoICL-Embodied-Sim50",
82
+ "revision": "788163e9cbb7a9f90d6ab4c6ed1f9e32289a2db4",
83
+ "path": "media/E43/E43_A__head.mp4",
84
+ "sha256": "4536b26ebd38a48591d3798c8c5d2f039c27d7791ce3f129f4d8c0f9f79e21e0"
85
+ },
86
+ "video_stream_preserved": true,
87
+ "transformation": "copy_video_stream_audio_removed"
88
+ },
89
+ "sim_status": "available",
90
+ "scope": "author_task_construction_materials",
91
+ "seed_and_environment_specific_evaluation_inputs": "benchmark/assignments.jsonl"
92
+ }