Add files using upload-large-folder tool

.gitattributes

@@ -59,3 +59,6 @@ assets/raw-sample-preview/fisheye_cam1_preview.mp4 filter=lfs diff=lfs merge=lfs
 assets/raw-sample-preview/stereo_right_preview.mp4 filter=lfs diff=lfs merge=lfs -text
 assets/raw-sample-preview/stereo_left_preview.mp4 filter=lfs diff=lfs merge=lfs -text
 assets/raw-sample-preview/fisheye_cam2_preview.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/foundation-pipelines/human-video-world-model-pipeline.png filter=lfs diff=lfs merge=lfs -text
+assets/foundation-pipelines/vision-language-action-pipeline.png filter=lfs diff=lfs merge=lfs -text
+assets/foundation-pipelines/spatial-intelligence-pipeline.png filter=lfs diff=lfs merge=lfs -text

assets/foundation-pipelines/README.md

@@ -0,0 +1,16 @@
+# Foundation Pipeline Placeholder Figures
+
+These three bitmap figures are ChatGPT image-generated placeholder visuals for
+the foundation pipeline tracks documented in `THREE_FOUNDATION_PIPELINES.md`
+and `docs/data/three_foundation_pipelines.json`.
+
+They are **pipeline placeholders**, not evidence of completed foundation-model
+training. Exact technical claims live in the surrounding Markdown, JSON, and
+website labels.
+
+| Track | Asset |
+| --- | --- |
+| Spatial intelligence models | `spatial-intelligence-pipeline.png` |
+| Human-video world models | `human-video-world-model-pipeline.png` |
+| Vision-language-action models | `vision-language-action-pipeline.png` |
+

assets/foundation-pipelines/prompts.md

@@ -0,0 +1,39 @@
+# ChatGPT Image Prompts
+
+## Spatial Intelligence
+
+Use case: infographic-diagram. Asset type: 16:9 website figure for Ropedia
+Xperience-10M foundation pipeline track. Create a polished text-free diagram
+image for a spatial intelligence model training pipeline. Show multi-view video
+frames and depth/pose streams flowing into a scene-object memory module, then
+spatial reasoning outputs like 3D structure, object permanence, counting, and
+question answering. Use a premium dark research-product presentation style,
+high contrast, crisp geometric panels, subtle neon green/cyan/white accents,
+clean technical linework, no decorative blobs, no logos, no readable text, no
+watermark.
+
+## Human-Video World Models
+
+Use case: infographic-diagram. Asset type: 16:9 website figure for Ropedia
+Xperience-10M foundation pipeline track. Create a polished text-free diagram
+image for a human-video world model training pipeline. Show observed egocentric
+video/audio/sensor windows flowing into a latent world-state model, then
+predicted future frames, future action bars, object/contact state changes, and
+uncertainty bands. Use a premium dark research-product presentation style,
+high contrast, crisp geometric panels, subtle neon green/teal/white accents
+with small amber highlights, clean technical linework, no decorative blobs, no
+logos, no readable text, no watermark.
+
+## Vision-Language-Action
+
+Use case: infographic-diagram. Asset type: 16:9 website figure for Ropedia
+Xperience-10M foundation pipeline track. Create a polished text-free diagram
+image for a vision-language-action model training pipeline. Show egocentric
+video frames, language caption tokens, hand/body motion traces, object/contact
+cues, and procedure labels flowing into a multimodal action policy module, then
+predicted action chunks, hand trajectory curves, contact decisions, and policy
+evaluation panels. Use a premium dark research-product presentation style,
+high contrast, crisp geometric panels, subtle neon green/cyan/white accents
+with small magenta highlights, clean technical linework, no decorative blobs,
+no logos, no readable text, no watermark.
+

docs/data/artifact_index.json

@@ -1,11 +1,12 @@
 {
   "title": "Ropedia Xperience-10M Task Suite Artifact Index",
-  "generated_at_utc": "2026-06-17T13:02:22+00:00",
+  "generated_at_utc": "2026-06-17T15:16:18+00:00",
   "status": "pass",
-  "artifact_count": 201,
+  "artifact_count": 204,
   "missing": [],
   "by_kind": {
     "project_path": 16,
+    "visual_asset": 3,
     "scaleup_contract": 7,
     "scaleup_status": 44,
     "publication_workflow": 6,
@@ -134,8 +135,8 @@
       "surface": "repo_hf",
       "shows": "Frames spatial intelligence, human-video world modeling, and vision-language-action as three pipeline tracks with explicit inputs, outputs, maturity, and next evidence gates.",
       "exists": true,
-      "bytes": 6538,
-      "sha256": "046c2873821962ab6d06fd21f8a2310358f9038b5bf543215fb962d850b92bab"
+      "bytes": 7437,
+      "sha256": "281a6349a7fd141460d7f911f0d80a841a38c99456363d1ffd6372cd94ca14b0"
     },
     {
       "id": "three_foundation_pipelines_json",
@@ -145,8 +146,41 @@
       "surface": "website_hf",
       "shows": "Machine-readable pipeline-track contract for the website and Hugging Face mirrors.",
       "exists": true,
-      "bytes": 5042,
-      "sha256": "3118aede3a63be4262839c9056278663e829e3a0a1deeb8a81c1f8ac977c1875"
+      "bytes": 6518,
+      "sha256": "e337901e7ddd2f8845987d4c41d9362e5fc780d3cb0659494576b7a0da53fb49"
+    },
+    {
+      "id": "spatial_intelligence_pipeline_placeholder",
+      "title": "Spatial intelligence pipeline placeholder",
+      "path": "docs/assets/foundation-pipelines/spatial-intelligence-pipeline.png",
+      "kind": "visual_asset",
+      "surface": "website_hf",
+      "shows": "ChatGPT image-generated placeholder visual for the spatial intelligence model training pipeline.",
+      "exists": true,
+      "bytes": 2337155,
+      "sha256": "ca98e2f5171497f6b97627ee8d0dee68f4aa929a2ba205e8b8e64e89f7f66f06"
+    },
+    {
+      "id": "human_video_world_model_pipeline_placeholder",
+      "title": "Human-video world model pipeline placeholder",
+      "path": "docs/assets/foundation-pipelines/human-video-world-model-pipeline.png",
+      "kind": "visual_asset",
+      "surface": "website_hf",
+      "shows": "ChatGPT image-generated placeholder visual for the human-video world-model training pipeline.",
+      "exists": true,
+      "bytes": 2356312,
+      "sha256": "cee2d717a97b88d8f5bae3e58fe202791a4f3073e488cb666acb0214117b735b"
+    },
+    {
+      "id": "vision_language_action_pipeline_placeholder",
+      "title": "Vision-language-action pipeline placeholder",
+      "path": "docs/assets/foundation-pipelines/vision-language-action-pipeline.png",
+      "kind": "visual_asset",
+      "surface": "website_hf",
+      "shows": "ChatGPT image-generated placeholder visual for the vision-language-action training pipeline.",
+      "exists": true,
+      "bytes": 2421011,
+      "sha256": "f8554f7df26ab79fef348740ce45ac3da032cb4085c490d62910ad1147dd1ecf"
     },
     {
       "id": "omni_model_extension_contract",
@@ -487,7 +521,7 @@
       "shows": "Machine-readable source-alignment pass/fail check for repo, website, and HF surfaces.",
       "exists": true,
       "bytes": 4432,
-      "sha256": "0c3ac97fdf21dfa3799a1aec3a007c436145f8ff6d70dc29f1eb002d1757c053"
+      "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
     },
     {
       "id": "source_alignment_validator",
@@ -608,7 +642,7 @@
       "shows": "Stores normalized 20-axis radar values, raw task metrics, Qwen3/Cosmos overlay mappings, branch-card caveats, and explicit scoreless status records.",
       "exists": true,
       "bytes": 231240,
-      "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+      "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
     },
     {
       "id": "single_episode_task_model_radar_json",
@@ -619,7 +653,7 @@
       "shows": "Machine-readable split radar for the one-episode Minimal and Neural MLP baselines, both scored on all 20 task contracts.",
       "exists": true,
       "bytes": 50973,
-      "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+      "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
     },
     {
       "id": "episode128_task_model_radar_json",
@@ -630,7 +664,7 @@
       "shows": "Machine-readable split radar for selected 128-episode metadata/raw baselines and verified Qwen3/Cosmos branches, preserving explicit scoreless cells.",
       "exists": true,
       "bytes": 187388,
-      "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+      "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
     },
     {
       "id": "task_method_20_result_matrix_json",
@@ -641,7 +675,7 @@
       "shows": "Machine-readable 9-method by 20-task matrix where every method has 20 records and scoreless cells carry unsupported/not-evaluated reasons.",
       "exists": true,
       "bytes": 129749,
-      "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+      "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
     },
     {
       "id": "task_method_20_result_matrix",
@@ -663,7 +697,7 @@
       "shows": "Machine-readable 180-record gap ledger with numeric scores, scoreless cells, explicit status reasons, and next evidence needed before new scores can be published.",
       "exists": true,
       "bytes": 55745,
-      "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+      "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
     },
     {
       "id": "task_method_20_gap_audit",
@@ -674,7 +708,7 @@
       "shows": "Reader-facing ledger that lists every scoreless method-task cell and the concrete target or model-output evidence required before it can become numeric.",
       "exists": true,
       "bytes": 15690,
-      "sha256": "2585cbb6ffbb87cff3e16347f93947930c7059f1053d68b6f98a4c70ecc58654"
+      "sha256": "bec8510557fee7505f68d697590eefdcaad96d70d9d9b201fab7a9bdc361a2ac"
     },
     {
       "id": "unified_task_model_radar_chart",
@@ -717,8 +751,8 @@
       "surface": "repo_hf",
       "shows": "Regenerates the direction-aware radar chart and machine-readable metric overlay JSON.",
       "exists": true,
-      "bytes": 50651,
-      "sha256": "c9c88870b9b6e72f290ea2f8ad5378dfb90f4aef5278f597f9a8d69c748ab352"
+      "bytes": 51243,
+      "sha256": "e0f995a01e8589a7f819dc5b766156c26e8b14e4db9c3c0c5e08be7a29b4de56"
     },
     {
       "id": "task_method_20_gap_audit_builder",
@@ -926,8 +960,8 @@
       "surface": "repo_hf",
       "shows": "Regenerates visual-asset hashes, dimensions, and source-script provenance.",
       "exists": true,
-      "bytes": 15412,
-      "sha256": "12eb9fa001f911f90d228362bb67f0fbc4a99be46180147d85a65742190dfd13"
+      "bytes": 16864,
+      "sha256": "df362654a5c65d7adedc924f2af93e1fbd248fa861f20aea1576473daaeb0b0d"
     },
     {
       "id": "brand_assets_json",
@@ -1107,8 +1141,8 @@
       "surface": "repo",
       "shows": "Fetches the published GitHub/HF URLs and compares live hashes and public-card markers against the release assets.",
       "exists": true,
-      "bytes": 56578,
-      "sha256": "d79de14ace608fbefc4cd0dac4f0851161882ef92749df78a3bd00c32b856fb0"
+      "bytes": 60253,
+      "sha256": "ad4b408e9e19339285e37e0c47bffac6a450ddd1a439bf11ab80a90cec27b1fb"
     },
     {
       "id": "reproducibility_contract",
@@ -1140,8 +1174,8 @@
       "surface": "repo_hf",
       "shows": "Generates the selective artifact catalog from local files.",
       "exists": true,
-      "bytes": 58004,
-      "sha256": "ffbbe7e9ab74fbb8b12dbd27968c31fd0e608cad2272329e1505039f57452b99"
+      "bytes": 59218,
+      "sha256": "38985fa362861b3975240ec62cc186378c84a8d0e11727651dc3cd2a87bfdd11"
     },
     {
       "id": "publication_audit",
@@ -1176,7 +1210,7 @@
       "volatile": true,
       "shows": "Confirms prepared GitHub/HF Space/artifact/model mirrors share the same critical data, figure, website HTML, and validator files.",
       "exists": true,
-      "bytes": 899853,
+      "bytes": 902747,
       "hash_policy": "existence_and_size_only"
     },
     {
@@ -1188,7 +1222,7 @@
       "volatile": true,
       "shows": "Confirms local website links, anchors, JSON data files, and referenced images resolve.",
       "exists": true,
-      "bytes": 18933,
+      "bytes": 19052,
       "hash_policy": "existence_and_size_only"
     },
     {

docs/data/mirror_parity.json

@@ -1,6 +1,6 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:03:46+00:00",
+  "generated_at_utc": "2026-06-17T13:55:47+00:00",
   "hf_root": "hf_publish",
   "summary": {
     "group_count": 611,
@@ -139,44 +139,44 @@
         "path": "repo:docs/data/artifact_index.json",
         "exists": true,
         "bytes": 109674,
-        "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+        "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         },
         "hf_model_data": {
           "path": "hf_model:data/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         },
         "hf_model": {
           "path": "hf_model:metrics/artifact_index.json",
           "exists": true,
           "bytes": 109674,
-          "sha256": "6940598729b16d07716133670f4c0f21c6203116f1f933f8fba8df0a549f665f"
+          "sha256": "9bec12c02579b9a14296a6f88f6fa2dcfb339d730f0d1068d9e55a7015bfbcc5"
         }
       },
       "failures": []
@@ -825,44 +825,44 @@
         "path": "repo:docs/data/publication_audit.json",
         "exists": true,
         "bytes": 8299,
-        "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+        "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         },
         "hf_model_data": {
           "path": "hf_model:data/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         },
         "hf_model": {
           "path": "hf_model:metrics/publication_audit.json",
           "exists": true,
           "bytes": 8299,
-          "sha256": "a68baa7a9b8afb067370c73b9a31ea4384b79d03acd56a8369aee76b83ded70d"
+          "sha256": "3d34bd58cd7f7a682d2a3a37786eb21db051d87ebec28c561b117b2c2388cee4"
         }
       },
       "failures": []
@@ -874,44 +874,44 @@
         "path": "repo:docs/data/public_surface_qa.json",
         "exists": true,
         "bytes": 6146,
-        "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+        "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         },
         "hf_model_data": {
           "path": "hf_model:data/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         },
         "hf_model": {
           "path": "hf_model:metrics/public_surface_qa.json",
           "exists": true,
           "bytes": 6146,
-          "sha256": "1498bdcfc914bad630bba9dcf25d6dc3102b6482a8f853b3e714710d7fdd63f9"
+          "sha256": "3e4cc531cf1c69099ffdf65073af9afbad473f86ac4049e8078e71dee7427a3b"
         }
       },
       "failures": []
@@ -1560,44 +1560,44 @@
         "path": "repo:docs/data/source_alignment_audit.json",
         "exists": true,
         "bytes": 4432,
-        "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+        "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         },
         "hf_model_data": {
           "path": "hf_model:data/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         },
         "hf_model": {
           "path": "hf_model:metrics/source_alignment_audit.json",
           "exists": true,
           "bytes": 4432,
-          "sha256": "60483ba20f4a24b05cf18fce2f12f632336e562bd11b935fbf5b4ccb1e31f34c"
+          "sha256": "5d013c6820fb7f582b4b4b9a55f98de20168ea1947d4bea64e11d16dbd521428"
         }
       },
       "failures": []
@@ -1658,44 +1658,44 @@
         "path": "repo:docs/data/single_episode_task_model_radar.json",
         "exists": true,
         "bytes": 50973,
-        "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+        "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         },
         "hf_model_data": {
           "path": "hf_model:data/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         },
         "hf_model": {
           "path": "hf_model:metrics/single_episode_task_model_radar.json",
           "exists": true,
           "bytes": 50973,
-          "sha256": "fb1beb92720407bfe51933a064b2bb9f28d1e45fb17af0e673827d255ee92fab"
+          "sha256": "07a6a8026e48e60d6a1ee0686d615645590ac3d95cc938fc9f0b26cbdea5d3a6"
         }
       },
       "failures": []
@@ -1707,44 +1707,44 @@
         "path": "repo:docs/data/episode128_task_model_radar.json",
         "exists": true,
         "bytes": 187388,
-        "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+        "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         },
         "hf_model_data": {
           "path": "hf_model:data/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         },
         "hf_model": {
           "path": "hf_model:metrics/episode128_task_model_radar.json",
           "exists": true,
           "bytes": 187388,
-          "sha256": "a3b71765b2435f085a7bb0492e05dbfc4c3c9b06cb65a4ec326160abb31f6ad8"
+          "sha256": "47e37a1b6bbbb3df98630dfab0de8e39e2c170400d1bce52054967a136dbc58c"
         }
       },
       "failures": []
@@ -1903,44 +1903,44 @@
         "path": "repo:docs/data/task_surface_integrity.json",
         "exists": true,
         "bytes": 45779,
-        "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+        "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         },
         "hf_model_data": {
           "path": "hf_model:data/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         },
         "hf_model": {
           "path": "hf_model:metrics/task_surface_integrity.json",
           "exists": true,
           "bytes": 45779,
-          "sha256": "7341690cd2e26af16c6b79a63304a988bac6c7264cda8f7cf1d51ff969226c66"
+          "sha256": "bf1b292db388f5a513100369078098b88705445908d8065a2b7907c584e40393"
         }
       },
       "failures": []
@@ -2001,44 +2001,44 @@
         "path": "repo:docs/data/task_method_20_result_matrix.json",
         "exists": true,
         "bytes": 129749,
-        "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+        "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         },
         "hf_model_data": {
           "path": "hf_model:data/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         },
         "hf_model": {
           "path": "hf_model:metrics/task_method_20_result_matrix.json",
           "exists": true,
           "bytes": 129749,
-          "sha256": "985d210609e40f06cb7d3d2819cd5e9330671a4506b1fb67c1917dcf4e2ea6e9"
+          "sha256": "58636609d9145bce26857ddee8e0fe4751ebee8429d4bef60fbe9d9daf7d2bd4"
         }
       },
       "failures": []
@@ -2050,44 +2050,44 @@
         "path": "repo:docs/data/task_method_20_gap_audit.json",
         "exists": true,
         "bytes": 55745,
-        "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+        "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         },
         "hf_model_data": {
           "path": "hf_model:data/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         },
         "hf_model": {
           "path": "hf_model:metrics/task_method_20_gap_audit.json",
           "exists": true,
           "bytes": 55745,
-          "sha256": "0049d8f850af3b47c100effde7d65940b63c3af2f91a5f29016bc3b066866576"
+          "sha256": "7cc10a067d029ae4d55869b2db1181e01fb5063ec5637111255a4f3d79dbb082"
         }
       },
       "failures": []
@@ -2148,44 +2148,44 @@
         "path": "repo:docs/data/unified_task_model_radar.json",
         "exists": true,
         "bytes": 231240,
-        "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+        "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         },
         "hf_model_data": {
           "path": "hf_model:data/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         },
         "hf_model": {
           "path": "hf_model:metrics/unified_task_model_radar.json",
           "exists": true,
           "bytes": 231240,
-          "sha256": "394d494ab99e5d978415764ed785521c038881dd678f3036783be70fef068d25"
+          "sha256": "87eb194c326323167b356448678fc9e2cc4b39610c48e6e14d368d55261d2745"
         }
       },
       "failures": []
@@ -2197,44 +2197,44 @@
         "path": "repo:docs/data/website_integrity.json",
         "exists": true,
         "bytes": 19052,
-        "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+        "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         },
         "hf_artifacts_data": {
           "path": "hf_artifacts:data/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         },
         "hf_model_data": {
           "path": "hf_model:data/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         },
         "hf_model_docs_data": {
           "path": "hf_model:docs/data/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         },
         "hf_model": {
           "path": "hf_model:metrics/website_integrity.json",
           "exists": true,
           "bytes": 19052,
-          "sha256": "2072e1c10faa6f73db575c69703e3d80fd42a59dd5af28f75003e8134e0e686e"
+          "sha256": "1be490f4c58971d19e1f9c614f40cbd64a776b8dec350438dae455596dfc182e"
         }
       },
       "failures": []
@@ -3430,21 +3430,21 @@
       "local": {
         "path": "repo:scripts/omni/collect_qwen3_future_task_probe_results.sh",
         "exists": true,
-        "bytes": 3346,
-        "sha256": "1220ef7c800500c3c91af5b3439396691fa470bfab645d9ff76102b9a27c2f0f"
+        "bytes": 3726,
+        "sha256": "35918a28d6e34acae6f71e667570354f82a1cdbd32816f603d248e19c356980c"
       },
       "mirrors": {
         "hf_artifacts": {
           "path": "hf_artifacts:scripts/omni/collect_qwen3_future_task_probe_results.sh",
           "exists": true,
-          "bytes": 3346,
-          "sha256": "1220ef7c800500c3c91af5b3439396691fa470bfab645d9ff76102b9a27c2f0f"
+          "bytes": 3726,
+          "sha256": "35918a28d6e34acae6f71e667570354f82a1cdbd32816f603d248e19c356980c"
         },
         "hf_model": {
           "path": "hf_model:scripts/omni/collect_qwen3_future_task_probe_results.sh",
           "exists": true,
-          "bytes": 3346,
-          "sha256": "1220ef7c800500c3c91af5b3439396691fa470bfab645d9ff76102b9a27c2f0f"
+          "bytes": 3726,
+          "sha256": "35918a28d6e34acae6f71e667570354f82a1cdbd32816f603d248e19c356980c"
         }
       },
       "failures": []
@@ -3530,21 +3530,21 @@
       "local": {
         "path": "repo:scripts/omni/eval_qwen3_omni_future_task_probes.py",
         "exists": true,
-        "bytes": 25313,
-        "sha256": "1b1e0b1f8bc215d5538168b5abafb705d8fe040fb6f1b2d4af33ec2775e0eebd"
+        "bytes": 32653,
+        "sha256": "5298a9c83252ac31cd30fa89e54834f98f6ccada8ffe10680f34773cbbe98d30"
       },
       "mirrors": {
         "hf_artifacts": {
           "path": "hf_artifacts:scripts/omni/eval_qwen3_omni_future_task_probes.py",
           "exists": true,
-          "bytes": 25313,
-          "sha256": "1b1e0b1f8bc215d5538168b5abafb705d8fe040fb6f1b2d4af33ec2775e0eebd"
+          "bytes": 32653,
+          "sha256": "5298a9c83252ac31cd30fa89e54834f98f6ccada8ffe10680f34773cbbe98d30"
         },
         "hf_model": {
           "path": "hf_model:scripts/omni/eval_qwen3_omni_future_task_probes.py",
           "exists": true,
-          "bytes": 25313,
-          "sha256": "1b1e0b1f8bc215d5538168b5abafb705d8fe040fb6f1b2d4af33ec2775e0eebd"
+          "bytes": 32653,
+          "sha256": "5298a9c83252ac31cd30fa89e54834f98f6ccada8ffe10680f34773cbbe98d30"
         }
       },
       "failures": []
@@ -4280,21 +4280,21 @@
       "local": {
         "path": "repo:scripts/build_unified_task_model_radar.py",
         "exists": true,
-        "bytes": 50651,
-        "sha256": "c9c88870b9b6e72f290ea2f8ad5378dfb90f4aef5278f597f9a8d69c748ab352"
+        "bytes": 51243,
+        "sha256": "e0f995a01e8589a7f819dc5b766156c26e8b14e4db9c3c0c5e08be7a29b4de56"
       },
       "mirrors": {
         "hf_artifacts": {
           "path": "hf_artifacts:scripts/build_unified_task_model_radar.py",
           "exists": true,
-          "bytes": 50651,
-          "sha256": "c9c88870b9b6e72f290ea2f8ad5378dfb90f4aef5278f597f9a8d69c748ab352"
+          "bytes": 51243,
+          "sha256": "e0f995a01e8589a7f819dc5b766156c26e8b14e4db9c3c0c5e08be7a29b4de56"
         },
         "hf_model": {
           "path": "hf_model:scripts/build_unified_task_model_radar.py",
           "exists": true,
-          "bytes": 50651,
-          "sha256": "c9c88870b9b6e72f290ea2f8ad5378dfb90f4aef5278f597f9a8d69c748ab352"
+          "bytes": 51243,
+          "sha256": "e0f995a01e8589a7f819dc5b766156c26e8b14e4db9c3c0c5e08be7a29b4de56"
         }
       },
       "failures": []
@@ -4330,21 +4330,21 @@
       "local": {
         "path": "repo:scripts/verify_live_publication.py",
         "exists": true,
-        "bytes": 56578,
-        "sha256": "d79de14ace608fbefc4cd0dac4f0851161882ef92749df78a3bd00c32b856fb0"
+        "bytes": 57383,
+        "sha256": "4cf40aa266827832734791b63862174a1d08a086bd97166fab31707320d5609c"
       },
       "mirrors": {
         "hf_artifacts": {
           "path": "hf_artifacts:scripts/verify_live_publication.py",
           "exists": true,
-          "bytes": 56578,
-          "sha256": "d79de14ace608fbefc4cd0dac4f0851161882ef92749df78a3bd00c32b856fb0"
+          "bytes": 57383,
+          "sha256": "4cf40aa266827832734791b63862174a1d08a086bd97166fab31707320d5609c"
         },
         "hf_model": {
           "path": "hf_model:scripts/verify_live_publication.py",
           "exists": true,
-          "bytes": 56578,
-          "sha256": "d79de14ace608fbefc4cd0dac4f0851161882ef92749df78a3bd00c32b856fb0"
+          "bytes": 57383,
+          "sha256": "4cf40aa266827832734791b63862174a1d08a086bd97166fab31707320d5609c"
         }
       },
       "failures": []
@@ -19545,26 +19545,26 @@
         "path": "repo:TASK_METHOD_20_GAP_AUDIT.md",
         "exists": true,
         "bytes": 15690,
-        "sha256": "2585cbb6ffbb87cff3e16347f93947930c7059f1053d68b6f98a4c70ecc58654"
+        "sha256": "bec8510557fee7505f68d697590eefdcaad96d70d9d9b201fab7a9bdc361a2ac"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:TASK_METHOD_20_GAP_AUDIT.md",
           "exists": true,
           "bytes": 15690,
-          "sha256": "2585cbb6ffbb87cff3e16347f93947930c7059f1053d68b6f98a4c70ecc58654"
+          "sha256": "bec8510557fee7505f68d697590eefdcaad96d70d9d9b201fab7a9bdc361a2ac"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:TASK_METHOD_20_GAP_AUDIT.md",
           "exists": true,
           "bytes": 15690,
-          "sha256": "2585cbb6ffbb87cff3e16347f93947930c7059f1053d68b6f98a4c70ecc58654"
+          "sha256": "bec8510557fee7505f68d697590eefdcaad96d70d9d9b201fab7a9bdc361a2ac"
         },
         "hf_model": {
           "path": "hf_model:TASK_METHOD_20_GAP_AUDIT.md",
           "exists": true,
           "bytes": 15690,
-          "sha256": "2585cbb6ffbb87cff3e16347f93947930c7059f1053d68b6f98a4c70ecc58654"
+          "sha256": "bec8510557fee7505f68d697590eefdcaad96d70d9d9b201fab7a9bdc361a2ac"
         }
       },
       "failures": []

docs/data/public_surface_qa.json

@@ -1,7 +1,7 @@
 {
   "title": "Ropedia Xperience-10M Public Project Surface",
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:02:21+00:00",
+  "generated_at_utc": "2026-06-17T15:16:15+00:00",
   "scope": "Repo README, GitHub Pages HTML, Hugging Face Space card, artifact dataset card, and model card.",
   "checks": [
     {
@@ -18,7 +18,7 @@
         "website_integrity": {
           "exists": true,
           "status": "pass",
-          "generated_at_utc": "2026-06-17T12:52:04+00:00"
+          "generated_at_utc": "2026-06-17T13:55:22+00:00"
         },
         "rendered_site_check": {
           "exists": true,
@@ -28,12 +28,12 @@
         "task_surface_integrity": {
           "exists": true,
           "status": "pass",
-          "generated_at_utc": "2026-06-17T07:22:34+00:00"
+          "generated_at_utc": "2026-06-17T13:55:20+00:00"
         },
         "source_alignment": {
           "exists": true,
           "status": "pass",
-          "generated_at_utc": "2026-06-17T12:51:54+00:00"
+          "generated_at_utc": "2026-06-17T13:55:20+00:00"
         },
         "scale_up_status": {
           "exists": true,
@@ -43,12 +43,12 @@
         "publication_package": {
           "exists": true,
           "status": "pass",
-          "generated_at_utc": "2026-06-17T12:52:11+00:00"
+          "generated_at_utc": "2026-06-17T13:55:30+00:00"
         },
         "mirror_parity": {
           "exists": true,
           "status": "pass",
-          "generated_at_utc": "2026-06-17T12:52:51+00:00"
+          "generated_at_utc": "2026-06-17T13:55:47+00:00"
         }
       },
       "failures": {}

docs/data/publication_audit.json

@@ -1,6 +1,6 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:03:53+00:00",
+  "generated_at_utc": "2026-06-17T15:17:33+00:00",
   "checks": [
     {
       "name": "required_publication_assets_present",
@@ -122,6 +122,9 @@
     "docs/assets/charts/unified_task_model_radar.svg": true,
     "docs/assets/charts/single_episode_task_model_radar.svg": true,
     "docs/assets/charts/episode128_task_model_radar.svg": true,
+    "docs/assets/foundation-pipelines/spatial-intelligence-pipeline.png": true,
+    "docs/assets/foundation-pipelines/human-video-world-model-pipeline.png": true,
+    "docs/assets/foundation-pipelines/vision-language-action-pipeline.png": true,
     "docs/assets/pipeline_diagram.png": true,
     "docs/assets/task_architectures.png": true,
     "results/episode_task_suite/summary_report.json": true,
@@ -200,8 +203,8 @@
     "github_repo": {
       "root": "repo",
       "exists": true,
-      "file_count": 1211,
-      "text_file_count": 1016,
+      "file_count": 1216,
+      "text_file_count": 1018,
       "largest_file": {
         "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
         "bytes": 55702978
@@ -222,7 +225,7 @@
     "hf_artifact_bundle": {
       "root": "hf_publish/artifacts",
       "exists": true,
-      "file_count": 2386,
+      "file_count": 2389,
       "text_file_count": 1036,
       "largest_file": {
         "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
@@ -233,7 +236,7 @@
     "hf_model_bundle": {
       "root": "hf_publish/model",
       "exists": true,
-      "file_count": 2820,
+      "file_count": 2824,
       "text_file_count": 1197,
       "largest_file": {
         "path": "pytorch_model.bin",

docs/data/single_episode_task_model_radar.json

@@ -1,7 +1,7 @@
 {
   "title": "Single-Episode 20-Task Radar",
   "status": "pass",
-  "generated_at_utc": "2026-06-17T07:04:16+00:00",
+  "generated_at_utc": "2026-06-17T13:55:02+00:00",
   "description": "Minimal and Neural MLP baselines on the one public sample episode, both scored on all 20 task contracts.",
   "task_count": 20,
   "method_count": 2,

docs/data/source_alignment_audit.json

@@ -1,7 +1,7 @@
 {
   "title": "Ropedia Xperience-10M Source Alignment Note",
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:02:52+00:00",
+  "generated_at_utc": "2026-06-17T15:17:20+00:00",
   "alignment_json": "docs/data/xperience10m_dataset_card_alignment.json",
   "alignment_summary": {
     "full_dataset_repo": "ropedia-ai/xperience-10m",

docs/data/task_method_20_gap_audit.json

@@ -1,5 +1,5 @@
 {
-  "generated_at_utc": "2026-06-17T07:04:39+00:00",
+  "generated_at_utc": "2026-06-17T13:55:12+00:00",
   "immediate_actions": [
     {
       "artifact": "docs/data/task_method_20_gap_audit.json",

docs/data/task_method_20_result_matrix.json

@@ -1,7 +1,7 @@
 {
   "title": "Task Method 20-Result Matrix",
   "status": "pass",
-  "generated_at_utc": "2026-06-17T07:04:16+00:00",
+  "generated_at_utc": "2026-06-17T13:55:02+00:00",
   "task_count": 20,
   "method_count": 9,
   "method_task_record_count": 180,

docs/data/task_surface_integrity.json

@@ -1,6 +1,6 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:02:52+00:00",
+  "generated_at_utc": "2026-06-17T15:17:20+00:00",
   "summary": {
     "task_count": 12,
     "expected_task_count": 12,

docs/data/three_foundation_pipelines.json

@@ -3,6 +3,13 @@
   "status": "pipeline_plan",
   "source_document": "THREE_FOUNDATION_PIPELINES.md",
   "claim_boundary": "These are supported pipeline directions, not three completed model-quality claims.",
+  "placeholder_assets": {
+    "status": "published_placeholders",
+    "asset_root": "docs/assets/foundation-pipelines",
+    "source": "ChatGPT image generation with repo-local prompt notes",
+    "source_prompt_file": "docs/assets/foundation-pipelines/prompts.md",
+    "note": "Images are visual placeholders for pipeline tracks. Technical claims remain governed by the Markdown/JSON contracts and verified metrics."
+  },
   "shared_principles": [
     "Use episode-level train/validation/test separation.",
     "Build manifest-first exporters before training.",
@@ -44,6 +51,9 @@
       "first_pipeline": "Build a spatial-memory exporter, start with metric depth and pose consistency tasks, then evaluate spatial QA, object permanence, counting, retrieval, and pose-aware consistency.",
       "current_maturity": "Ready as a pipeline and evaluation contract.",
       "next_gate": "Raw depth and pose artifacts plus held-out multi-episode spatial metrics.",
+      "placeholder_image": "docs/assets/foundation-pipelines/spatial-intelligence-pipeline.png",
+      "website_image": "assets/foundation-pipelines/spatial-intelligence-pipeline.png",
+      "image_alt": "Placeholder visual for the spatial intelligence pipeline: multiview video, depth, and pose inputs feeding scene memory and spatial reasoning outputs.",
       "avoid_claiming_now": [
         "full neural rendering",
         "full 3D reconstruction",
@@ -82,6 +92,9 @@
       "first_pipeline": "Keep Qwen-style structured future probes for task interpretability, keep Cosmos-style dynamics branches separate, and add latent or feature-reconstruction metrics before claiming world-model quality.",
       "current_maturity": "Partially evidenced by current future-task probes and Cosmos-style branch artifacts.",
       "next_gate": "Stronger future-state metrics, qualitative future examples, and held-out episode breakdowns.",
+      "placeholder_image": "docs/assets/foundation-pipelines/human-video-world-model-pipeline.png",
+      "website_image": "assets/foundation-pipelines/human-video-world-model-pipeline.png",
+      "image_alt": "Placeholder visual for the human-video world model pipeline: observed interaction windows feeding temporal dynamics and future-state outputs.",
       "avoid_claiming_now": [
         "strong world model from structured future-task scores alone",
         "visual future quality without visual or latent future metrics"
@@ -118,6 +131,9 @@
       "first_pipeline": "Define the action space, use existing 20-task next-action/contact/object-conditioned tasks first, then add hand-trajectory or policy-compatible action chunks after conversion is traceable.",
       "current_maturity": "Feasible but gated by action-target conversion.",
       "next_gate": "Traceable action tokens, normalization, retargeting metadata, and held-out policy metrics.",
+      "placeholder_image": "docs/assets/foundation-pipelines/vision-language-action-pipeline.png",
+      "website_image": "assets/foundation-pipelines/vision-language-action-pipeline.png",
+      "image_alt": "Placeholder visual for the vision-language-action pipeline: video, language, motion, and contact cues feeding action-chunk outputs.",
       "avoid_claiming_now": [
         "robot policy quality",
         "policy generalization before action-space evidence exists"

docs/data/unified_task_model_radar.json

@@ -1,7 +1,7 @@
 {
   "title": "Unified 20-Task Model Radar",
   "status": "pass",
-  "generated_at_utc": "2026-06-17T07:04:16+00:00",
+  "generated_at_utc": "2026-06-17T13:55:02+00:00",
   "task_count": 20,
   "method_count": 9,
   "method_task_record_count": 180,

docs/data/website_integrity.json

@@ -1,14 +1,14 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-17T13:02:53+00:00",
+  "generated_at_utc": "2026-06-17T15:17:21+00:00",
   "docs_root": "docs",
   "site_base": "/ropedia-xperience-10m-task-suite/",
   "summary": {
     "html_pages": 4,
-    "local_references": 181,
+    "local_references": 187,
     "external_reference_count": 123,
     "json_files": 47,
-    "image_assets_referenced": 25,
+    "image_assets_referenced": 28,
     "failure_count": 0
   },
   "failures": {
@@ -80,8 +80,8 @@
       "name": "project_overview_precedes_progress_ledger",
       "status": "pass",
       "reason": "The project overview should appear before the deeper progress ledger.",
-      "overview_index": 87197,
-      "evidence_index": 115940
+      "overview_index": 88890,
+      "evidence_index": 119623
     },
     {
       "name": "project_status_links_json",
@@ -159,9 +159,9 @@
       "name": "evaluation_protocol_between_overview_and_progress",
       "status": "pass",
       "reason": "The evaluation protocol should appear before the deeper evidence ledger.",
-      "overview_index": 87197,
-      "protocol_index": 112121,
-      "evidence_index": 115940
+      "overview_index": 88890,
+      "protocol_index": 115804,
+      "evidence_index": 119623
     },
     {
       "name": "evaluation_protocol_links_json",
@@ -277,8 +277,8 @@
     {
       "path": "index.html",
       "id_count": 90,
-      "reference_count": 157,
-      "image_count": 31
+      "reference_count": 163,
+      "image_count": 34
     },
     {
       "path": "research_roadmap.html",
@@ -301,7 +301,7 @@
     },
     {
       "path": "data/artifact_index.json",
-      "bytes": 109674,
+      "bytes": 111262,
       "top_level_type": "dict"
     },
     {
@@ -331,7 +331,7 @@
     },
     {
       "path": "data/figure_index.json",
-      "bytes": 17287,
+      "bytes": 19501,
       "top_level_type": "dict"
     },
     {
@@ -346,7 +346,7 @@
     },
     {
       "path": "data/mirror_parity.json",
-      "bytes": 899853,
+      "bytes": 902747,
       "top_level_type": "dict"
     },
     {
@@ -506,7 +506,7 @@
     },
     {
       "path": "data/three_foundation_pipelines.json",
-      "bytes": 5042,
+      "bytes": 6518,
       "top_level_type": "dict"
     },
     {
@@ -521,7 +521,7 @@
     },
     {
       "path": "data/website_integrity.json",
-      "bytes": 18933,
+      "bytes": 19052,
       "top_level_type": "dict"
     },
     {
@@ -630,6 +630,30 @@
       "format": "SVG",
       "has_viewbox": true
     },
+    {
+      "path": "assets/foundation-pipelines/human-video-world-model-pipeline.png",
+      "exists": true,
+      "bytes": 2356312,
+      "width": 1672,
+      "height": 941,
+      "format": "PNG"
+    },
+    {
+      "path": "assets/foundation-pipelines/spatial-intelligence-pipeline.png",
+      "exists": true,
+      "bytes": 2337155,
+      "width": 1672,
+      "height": 941,
+      "format": "PNG"
+    },
+    {
+      "path": "assets/foundation-pipelines/vision-language-action-pipeline.png",
+      "exists": true,
+      "bytes": 2421011,
+      "width": 1672,
+      "height": 941,
+      "format": "PNG"
+    },
     {
       "path": "assets/modalities/audio.png",
       "exists": true,

scripts/omni/collect_qwen3_future_task_probe_results.sh

@@ -14,12 +14,9 @@ REMOTE_RUN_DIR="${REMOTE_ROOT}/${RESULT_ROOT}/${RUN_ID}"
 LOCAL_RUN_DIR="${PROJECT_ROOT}/${RESULT_ROOT}/${RUN_ID}"
 LOCAL_LAUNCHER_DIR="${PROJECT_ROOT}/${RESULT_ROOT}/deferred_launchers"
 REMOTE_LAUNCHER_LOG="${REMOTE_ROOT}/${RESULT_ROOT}/deferred_launchers/${RUN_ID}.launcher.log"
+TASKS_CSV="${TASKS_CSV:-long_horizon_next_action,next_subtask_forecast,object_set_forecast}"
 
-TASKS=(
-  long_horizon_next_action
-  next_subtask_forecast
-  object_set_forecast
-)
+IFS=',' read -r -a TASKS <<< "$TASKS_CSV"
 
 echo "checking remote run ${REMOTE_HOST}:${REMOTE_RUN_DIR}"
 ssh "$REMOTE_HOST" "cd '$REMOTE_ROOT' && test -s '${RESULT_ROOT}/${RUN_ID}/summary.json'"
@@ -33,19 +30,24 @@ ssh "$REMOTE_HOST" "test -s '$REMOTE_LAUNCHER_LOG'" >/dev/null 2>&1 \
   && rsync -av "${REMOTE_HOST}:${REMOTE_LAUNCHER_LOG}" "$LOCAL_LAUNCHER_DIR/" \
   || true
 
-python3 - "$PROJECT_ROOT" "$RUN_ID" <<'PY'
+python3 - "$PROJECT_ROOT" "$RUN_ID" "$TASKS_CSV" <<'PY'
 import json
 import sys
 from pathlib import Path
 
 root = Path(sys.argv[1])
 run_id = sys.argv[2]
+task_ids = [item.strip() for item in sys.argv[3].split(",") if item.strip()]
 run_dir = root / "results/omni_finetune" / run_id
-expected = {
+metric_key_by_task = {
+    "temporal_order": "temporal_order_f1",
+    "misalignment_detection": "misalignment_detection_f1",
     "long_horizon_next_action": "long_horizon_next_action_macro_f1",
     "next_subtask_forecast": "next_subtask_forecast_macro_f1",
     "object_set_forecast": "object_set_forecast_micro_f1",
+    "time_to_transition": "time_to_transition_mae",
 }
+expected = {task_id: metric_key_by_task[task_id] for task_id in task_ids}
 
 summary_path = run_dir / "summary.json"
 if not summary_path.exists():

scripts/omni/eval_qwen3_omni_future_task_probes.py

@@ -1,14 +1,17 @@
 #!/usr/bin/env python3
 """Evaluate Qwen3-Omni on future-target task probes from the 128-episode JSON.
 
-This runner scores only task targets that can be derived from the current
-multi-episode JSON export:
+This runner scores task targets that can be derived from the current
+multi-episode JSON export and staged media:
 
 - Task 13: long-horizon next action, +100 frames.
 - Task 14: long-horizon next subtask, +100 frames.
 - Task 17: future object set, +100 frames.
+- Task 11: temporal order from two staged video windows.
+- Task 12: audio-video misalignment from staged video/audio windows.
+- Task 20: capped frames until next action transition.
 
-It does not fabricate scores for regression, retrieval, raw-caption, or
+It does not fabricate scores for retrieval, raw-caption, raw hand-pose, or
 missing-modality targets.
 """
 
@@ -16,7 +19,9 @@ from __future__ import annotations
 
 import argparse
 import csv
+import hashlib
 import json
+import re
 import time
 from collections import OrderedDict
 from pathlib import Path
@@ -37,6 +42,32 @@ from qwen3_omni_dataset_utils import (
 
 TASK_SPECS: OrderedDict[str, dict[str, Any]] = OrderedDict(
     [
+        (
+            "temporal_order",
+            {
+                "task_number": 11,
+                "label": "Temporal Order Verification",
+                "family": "classification",
+                "metric_key": "temporal_order_f1",
+                "prediction_key": "temporal_order",
+                "target_field": None,
+                "option_field": None,
+                "options": ["correct", "reversed"],
+            },
+        ),
+        (
+            "misalignment_detection",
+            {
+                "task_number": 12,
+                "label": "Multimodal Misalignment Detection",
+                "family": "classification",
+                "metric_key": "misalignment_detection_f1",
+                "prediction_key": "misalignment_detection",
+                "target_field": None,
+                "option_field": None,
+                "options": ["aligned", "shifted"],
+            },
+        ),
         (
             "long_horizon_next_action",
             {
@@ -73,6 +104,18 @@ TASK_SPECS: OrderedDict[str, dict[str, Any]] = OrderedDict(
                 "option_field": None,
             },
         ),
+        (
+            "time_to_transition",
+            {
+                "task_number": 20,
+                "label": "Time to Transition",
+                "family": "regression",
+                "metric_key": "time_to_transition_mae",
+                "prediction_key": "time_to_transition_frames",
+                "target_field": None,
+                "option_field": None,
+            },
+        ),
     ]
 )
 
@@ -207,6 +250,22 @@ def future_index_map(samples: list[dict[str, Any]], frame_offset: int) -> dict[i
     return mapping
 
 
+def time_to_transition_map(samples: list[dict[str, Any]], cap_frames: int = 200) -> dict[int, int]:
+    mapping: dict[int, int] = {}
+    for indices in by_episode_sorted(samples).values():
+        actions = [normalize_text(answer(samples[idx]).get("action")) for idx in indices]
+        starts = [row_start(samples[idx]) for idx in indices]
+        for pos, idx in enumerate(indices):
+            current_action = actions[pos]
+            target = cap_frames
+            for next_pos in range(pos + 1, len(indices)):
+                if actions[next_pos] and actions[next_pos] != current_action:
+                    target = min(cap_frames, max(0, starts[next_pos] - starts[pos]))
+                    break
+            mapping[idx] = target
+    return mapping
+
+
 def parse_json_object(text: str) -> dict[str, Any]:
     raw = str(text or "").strip()
     if raw.startswith("```"):
@@ -227,7 +286,25 @@ def parse_json_object(text: str) -> dict[str, Any]:
     return payload if isinstance(payload, dict) else {}
 
 
+def stable_variant(task_id: str, sample: dict[str, Any]) -> bool:
+    key = f"{task_id}::{sample.get('id')}"
+    digest = hashlib.sha1(key.encode("utf-8")).hexdigest()
+    return int(digest[:2], 16) % 2 == 0
+
+
+def media_video_path(sample: dict[str, Any]) -> str | None:
+    media = sample.get("media") if isinstance(sample.get("media"), dict) else {}
+    return media.get("mosaic_video_path") or sample.get("primary_video_path")
+
+
+def media_audio_path(sample: dict[str, Any]) -> str | None:
+    media = sample.get("media") if isinstance(sample.get("media"), dict) else {}
+    return media.get("audio_path")
+
+
 def task_options(sample: dict[str, Any], spec: dict[str, Any]) -> list[str]:
+    if isinstance(spec.get("options"), list):
+        return [str(item) for item in spec["options"]]
     option_field = spec.get("option_field")
     options = sample.get(option_field) if option_field else None
     if isinstance(options, list) and options:
@@ -247,8 +324,12 @@ def build_task_prompt(sample: dict[str, Any], future_sample: dict[str, Any], tas
         f"Task {spec['task_number']}: {spec['label']}",
         f"Episode: {sample.get('episode_id')}",
         f"Current visible/audio context frames: {start}-{end}",
-        f"Predict the target at the future window starting near frame {start + future_frames} (resolved target start frame {future_start}).",
     ]
+    if task_id in {"long_horizon_next_action", "next_subtask_forecast", "object_set_forecast"}:
+        lines.append(
+            f"Predict the target at the future window starting near frame {start + future_frames} "
+            f"(resolved target start frame {future_start})."
+        )
     options = task_options(sample, spec)
     if task_id == "long_horizon_next_action":
         lines.extend(
@@ -276,6 +357,35 @@ def build_task_prompt(sample: dict[str, Any], future_sample: dict[str, Any], tas
                 "List the objects likely to be active or manipulated in that future window. Use short object names.",
             ]
         )
+    elif task_id == "temporal_order":
+        lines.extend(
+            [
+                "You will receive two video clips named Clip A and Clip B.",
+                "Return JSON only with this schema:",
+                f'{{"{prediction_key}":"<correct or reversed>"}}',
+                "Answer correct if Clip A happens before Clip B in the same episode.",
+                "Answer reversed if Clip A happens after Clip B in the same episode.",
+            ]
+        )
+    elif task_id == "misalignment_detection":
+        lines.extend(
+            [
+                "You will receive one video clip and one audio clip.",
+                "Return JSON only with this schema:",
+                f'{{"{prediction_key}":"<aligned or shifted>"}}',
+                "Answer aligned if the audio belongs to the same time window as the video.",
+                "Answer shifted if the audio comes from a later shifted window in the same episode.",
+            ]
+        )
+    elif task_id == "time_to_transition":
+        lines.extend(
+            [
+                "Estimate how many frames remain until the next action-label boundary.",
+                "The answer is capped at 200 frames.",
+                "Return JSON only with this schema:",
+                f'{{"{prediction_key}":<integer from 0 to 200>}}',
+            ]
+        )
     else:
         raise ValueError(f"unknown task: {task_id}")
     return "\n".join(lines)
@@ -290,14 +400,30 @@ def build_messages(
     *,
     include_audio: bool = True,
 ) -> list[dict[str, Any]]:
-    media = sample.get("media") if isinstance(sample.get("media"), dict) else {}
-    video_path = media.get("mosaic_video_path") or sample.get("primary_video_path")
-    audio_path = media.get("audio_path")
+    video_path = media_video_path(sample)
+    audio_path = media_audio_path(sample)
     content: list[dict[str, Any]] = []
-    if video_path:
-        content.append({"type": "video", "video": video_path})
-    if include_audio and audio_path:
-        content.append({"type": "audio", "audio": audio_path})
+    if task_id == "temporal_order":
+        future_video_path = media_video_path(future_sample)
+        if stable_variant(task_id, sample):
+            first_video, second_video = video_path, future_video_path
+        else:
+            first_video, second_video = future_video_path, video_path
+        if first_video:
+            content.append({"type": "video", "video": first_video})
+        if second_video:
+            content.append({"type": "video", "video": second_video})
+    elif task_id == "misalignment_detection":
+        paired_audio_path = audio_path if stable_variant(task_id, sample) else media_audio_path(future_sample)
+        if video_path:
+            content.append({"type": "video", "video": video_path})
+        if include_audio and paired_audio_path:
+            content.append({"type": "audio", "audio": paired_audio_path})
+    else:
+        if video_path:
+            content.append({"type": "video", "video": video_path})
+        if include_audio and audio_path:
+            content.append({"type": "audio", "audio": audio_path})
     content.append({"type": "text", "text": build_task_prompt(sample, future_sample, task_id, spec, future_frames)})
     return [
         {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},
@@ -394,10 +520,32 @@ def extract_prediction(raw: str, sample: dict[str, Any], spec: dict[str, Any]) -
     value = payload.get(spec["prediction_key"])
     if spec["family"] == "multi_label":
         return normalize_objects(value)
+    if spec["family"] == "regression":
+        match = re.search(r"-?\d+(?:\.\d+)?", str(value if value is not None else raw))
+        if not match:
+            return None
+        return max(0.0, min(200.0, float(match.group(0))))
     options = task_options(sample, spec)
     return match_label(str(value or raw), options) if options else normalize_text(value)
 
 
+def task_target_value(
+    task_id: str,
+    sample: dict[str, Any],
+    future_sample: dict[str, Any],
+    spec: dict[str, Any],
+    transition_targets: dict[int, int],
+    sample_idx: int,
+) -> Any:
+    if task_id == "temporal_order":
+        return "correct" if stable_variant(task_id, sample) else "reversed"
+    if task_id == "misalignment_detection":
+        return "aligned" if stable_variant(task_id, sample) else "shifted"
+    if task_id == "time_to_transition":
+        return float(transition_targets[sample_idx])
+    return task_target(future_sample, spec)
+
+
 def object_set_metrics(rows: list[dict[str, Any]]) -> dict[str, float]:
     tp = fp = fn = exact = 0
     for row in rows:
@@ -419,6 +567,26 @@ def object_set_metrics(rows: list[dict[str, Any]]) -> dict[str, float]:
     }
 
 
+def regression_metrics(rows: list[dict[str, Any]]) -> dict[str, float]:
+    errors = []
+    within_20 = 0
+    for row in rows:
+        true_value = float(row.get("true_value") or 0.0)
+        pred_value = row.get("predicted_value")
+        if pred_value is None:
+            pred_value = 200.0
+        err = abs(float(pred_value) - true_value)
+        errors.append(err)
+        within_20 += int(err <= 20.0)
+    mae = float(np.mean(errors)) if errors else 0.0
+    return {
+        "num_samples": len(rows),
+        "mae": mae,
+        "time_to_transition_mae": mae,
+        "within_20_frames": within_20 / len(rows) if rows else 0.0,
+    }
+
+
 def score_task(task_id: str, spec: dict[str, Any], rows: list[dict[str, Any]], output_dir: Path, args: argparse.Namespace) -> dict[str, Any]:
     task_dir = output_dir / task_id
     task_dir.mkdir(parents=True, exist_ok=True)
@@ -471,11 +639,17 @@ def score_task(task_id: str, spec: dict[str, Any], rows: list[dict[str, Any]], o
         metrics[f"{task_id}_accuracy"] = metrics["accuracy"]
         write_csv(task_dir / "per_class_metrics.csv", per_class, ["class_name", "support", "predicted", "precision", "recall", "f1"])
         primary_score = metrics["macro_f1"]
-    else:
+    elif spec["family"] == "multi_label":
         metrics = object_set_metrics(rows)
         metrics[f"{task_id}_micro_f1"] = metrics["micro_f1"]
         metrics[f"{task_id}_exact_match"] = metrics["exact_match"]
         primary_score = metrics["micro_f1"]
+    elif spec["family"] == "regression":
+        metrics = regression_metrics(rows)
+        primary_score = metrics["mae"]
+    else:
+        raise ValueError(f"unsupported task family: {spec['family']}")
+    metrics[spec["metric_key"]] = primary_score
 
     metrics.update(
         {
@@ -516,6 +690,7 @@ def main() -> int:
     selected_tasks = select_tasks(args.tasks)
     samples = load_jsonl(args.dataset_jsonl)
     future_map = future_index_map(samples, args.future_frames)
+    transition_targets = time_to_transition_map(samples)
     eval_indices = [idx for idx in select_eval_indices(samples, args) if idx in future_map]
     if not eval_indices:
         raise ValueError("No evaluation samples with future targets selected.")
@@ -554,8 +729,14 @@ def main() -> int:
                 continue
             started = time.time()
             raw = generate_messages(model, processor, sample, future_sample, task_id, spec, args)
-            true_value = task_target(future_sample, spec)
+            true_value = task_target_value(task_id, sample, future_sample, spec, transition_targets, sample_idx)
             predicted_value = extract_prediction(raw, sample, spec)
+            if spec["family"] == "classification":
+                correct = int(true_value == predicted_value)
+            elif spec["family"] == "multi_label":
+                correct = int(set(true_value) == set(predicted_value))
+            else:
+                correct = int(predicted_value is not None and abs(float(true_value) - float(predicted_value)) <= 20.0)
             row = {
                 "prediction_id": pred_id,
                 "id": sample.get("id"),
@@ -571,7 +752,7 @@ def main() -> int:
                 "true_value": true_value,
                 "predicted_value": predicted_value,
                 "raw_prediction": raw,
-                "correct": int(true_value == predicted_value) if spec["family"] == "classification" else int(set(true_value) == set(predicted_value)),
+                "correct": correct,
             }
             partial_by_task[task_id][pred_id] = row
             append_jsonl(partial_path, row)