Keep dataset visual markers inline with headings and table labels

README.md

@@ -33,21 +33,21 @@ RoboValue provides expert demonstrations for **robotic value models**, covering
   <a href="#overview">Overview</a> · <a href="#simulation">Simulation</a> · <a href="#real-world">Real-World</a> · <a href="#repository-layout">Repository</a> · <a href="#download-and-reading">Download &amp; Read</a>
 </p>
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/overview.svg" alt="" width="24" height="24" /> 1. Dataset Overview <a name="overview"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/overview.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 1. Dataset Overview <a name="overview"></a>
 
 | Domain | Platform | Tasks | Full-Shot | One-Shot |
 | --- | --- | ---: | ---: | ---: |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> [**Simulation**](#simulation) | ARX | 15 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-1500.svg" alt="1,500" width="76" height="28" /> episodes | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-15.svg" alt="15" width="52" height="28" /> episodes |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> [**Real-World**](#real-world) | AgiBot Genie02 | 20 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-2000.svg" alt="2,000" width="76" height="28" /> episodes | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-20.svg" alt="20" width="52" height="28" /> episodes |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> [**Simulation**](#simulation) | ARX | 15 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-1500.svg" alt="1,500" width="76" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> episodes | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-15.svg" alt="15" width="52" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> episodes |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> [**Real-World**](#real-world) | AgiBot Genie02 | 20 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-2000.svg" alt="2,000" width="76" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> episodes | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-20.svg" alt="20" width="52" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> episodes |
 
 | Setting | Demonstrations | Use |
 | --- | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="16" height="16" /> **Full-Shot** | 100 per task | Training or fine-tuning on the complete expert collection |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="16" height="16" /> **One-Shot** | 1 per task | Reference conditioning or task-specific adaptation |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Full-Shot** | 100 per task | Training or fine-tuning on the complete expert collection |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **One-Shot** | 1 per task | Reference conditioning or task-specific adaptation |
 
 The two One-Shot subsets contain **35 episodes, 23,747 time steps, and 105 camera videos** in total. These are standard in-domain expert demonstrations; value or progress supervision is constructed by the training method.
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="24" height="24" /> 2. Simulation <a name="simulation"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 2. Simulation <a name="simulation"></a>
 
 ### 2.1 Overview
 
@@ -59,11 +59,11 @@ In both Full-Shot and One-Shot, an **episode** is a complete task demonstration
 
 | Content | Representation |
 | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" /> Robot state | `observation.state`: `float32[14]` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" /> Robot action | `action`: `float32[14]` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> RGB observations | High, left-wrist, and right-wrist views, each **640 × 480** |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" /> Instruction and identity | Task instruction, `task_index`, `episode_index`, `frame_index` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Time | `timestamp` in seconds; **25 FPS** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Robot state | `observation.state`: `float32[14]` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Robot action | `action`: `float32[14]` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> RGB observations | High, left-wrist, and right-wrist views, each **640 × 480** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Instruction and identity | Task instruction, `task_index`, `episode_index`, `frame_index` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time | `timestamp` in seconds; **25 FPS** |
 
 <details>
 <summary><strong>Simulation feature reference: states, actions, and cameras</strong></summary>
@@ -104,27 +104,27 @@ Camera keys are `observation.images.cam_high`, `observation.images.cam_left_wris
 
 </details>
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="20" height="20" /> 2.2 Full-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 2.2 Full-Shot
 
 **15 tasks × 100 demonstrations = 1,500 episodes.**
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> Training split |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Training split |
 | ---: | --- | ---: | ---: | --- |
 | **789,612** | **4,500 · AV1 / H.264** | **≈45.47 GB** | ≈8.77 hours | `train: 0:1500` |
 
 [**Browse files →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/simulation/full-shot) · [Download and read](#download-and-reading)
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="20" height="20" /> 2.3 One-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 2.3 One-Shot
 
 **15 tasks × 1 demonstration = 15 episodes.** Uses the same task definitions as Full-Shot.
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> Training split |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Training split |
 | ---: | --- | ---: | ---: | --- |
 | **8,080** | **45 · AV1** | **≈0.50 GB** | ≈5.39 minutes | `train: 0:15` |
 
 [**Browse files →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/simulation/one-shot) · [Task–episode–row mapping](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/blob/main/data/simulation/one-shot/provenance/episode_mapping.json) · [Download and read](#download-and-reading)
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="24" height="24" /> 3. Real-World <a name="real-world"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 3. Real-World <a name="real-world"></a>
 
 ### 3.1 Overview
 
@@ -136,11 +136,11 @@ In both Full-Shot and One-Shot, an **episode** is a complete task demonstration
 
 | Content | Representation |
 | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" /> Robot state | `observation.state`: `float32[290]`, including joint/end-effector state and auxiliary fields |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" /> Robot action | `action`: `float32[52]`, including joint targets, end-effector target poses, and other control fields |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> RGB observations | Head **640 × 400**; each wrist **640 × 528** |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" /> Instruction and identity | Task instruction, `task_index`, `episode_index`, `frame_index` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Time | `timestamp` in seconds; **30 FPS** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Robot state | `observation.state`: `float32[290]`, including joint/end-effector state and auxiliary fields |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Robot action | `action`: `float32[52]`, including joint targets, end-effector target poses, and other control fields |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> RGB observations | Head **640 × 400**; each wrist **640 × 528** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Instruction and identity | Task instruction, `task_index`, `episode_index`, `frame_index` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time | `timestamp` in seconds; **30 FPS** |
 
 Camera keys are `observation.images.cam_high`, `observation.images.cam_left_wrist`, and `observation.images.cam_right_wrist`. Loaded images use **CHW** order.
 
@@ -220,27 +220,27 @@ Indices match the training data's `task_index`. Display names and descriptions f
 
 </details>
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-real-world.svg" alt="" width="20" height="20" /> 3.2 Full-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-real-world.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 3.2 Full-Shot
 
 **20 tasks × 100 demonstrations = 2,000 episodes.**
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> Training split |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Training split |
 | ---: | --- | ---: | ---: | --- |
 | **1,530,987** | **6,000 · HEVC** | **≈23.23 GB** | ≈14.18 hours | `train: 0:2000` |
 
 [**Browse files →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/real-world/full-shot) · [Download and read](#download-and-reading)
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-real-world.svg" alt="" width="20" height="20" /> 3.3 One-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-real-world.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 3.3 One-Shot
 
 **20 tasks × 1 demonstration = 20 episodes.** Uses the same task definitions as Full-Shot.
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> Training split |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Videos | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Size | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Total duration | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Training split |
 | ---: | --- | ---: | ---: | --- |
 | **15,667** | **60 · HEVC** | **≈0.23 GB** | ≈8.70 minutes | `train: 0:20` |
 
 [**Browse files →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/real-world/one-shot) · [Task–episode–row mapping](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/blob/main/data/real-world/one-shot/provenance/episode_mapping.json) · [Download and read](#download-and-reading)
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="24" height="24" /> 4. Repository Layout <a name="repository-layout"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 4. Repository Layout <a name="repository-layout"></a>
 
 ```text
 README.md                         English dataset card
@@ -277,18 +277,18 @@ Data rows span `[dataset_from_index, dataset_to_index)`; video segments are reco
 
 </details>
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="24" height="24" /> 5. Download and Reading <a name="download-and-reading"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 5. Download and Reading <a name="download-and-reading"></a>
 
 ### 5.1 Choose a subset
 
 Select one `prefix` to download only that subset. Sizes use decimal GB.
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="16" height="16" /> Subset | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="16" height="16" /> `prefix` | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> Episodes / time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> Size |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Subset | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> `prefix` | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Episodes / time steps | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> Size |
 | --- | --- | ---: | ---: |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> **Simulation Full-Shot** | `data/simulation/full-shot` | 1,500 / 789,612 | ≈45.47 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> **Simulation One-Shot** | `data/simulation/one-shot` | 15 / 8,080 | ≈0.50 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> **Real-World Full-Shot** | `data/real-world/full-shot` | 2,000 / 1,530,987 | ≈23.23 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> **Real-World One-Shot** | `data/real-world/one-shot` | 20 / 15,667 | ≈0.23 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Simulation Full-Shot** | `data/simulation/full-shot` | 1,500 / 789,612 | ≈45.47 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Simulation One-Shot** | `data/simulation/one-shot` | 15 / 8,080 | ≈0.50 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Real-World Full-Shot** | `data/real-world/full-shot` | 2,000 / 1,530,987 | ≈23.23 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Real-World One-Shot** | `data/real-world/one-shot` | 20 / 15,667 | ≈0.23 GB |
 
 ### 5.2 Download and read
 

README_zh.md

@@ -16,21 +16,21 @@ RoboValue 提供面向**机器人价值模型**的专家示例，覆盖 **15 个
   <a href="#overview">数据总览</a> · <a href="#simulation">仿真世界</a> · <a href="#real-world">真实世界</a> · <a href="#repository-layout">仓库目录</a> · <a href="#download-and-reading">下载与读取</a>
 </p>
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/overview.svg" alt="" width="24" height="24" /> 1. 数据总览 <a name="overview"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/overview.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 1. 数据总览 <a name="overview"></a>
 
 | 领域 | 采集平台 | 任务数 | Full-Shot | One-Shot |
 | --- | --- | ---: | ---: | ---: |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> [**仿真世界**](#simulation) | ARX | 15 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-1500.svg" alt="1,500" width="76" height="28" /> 条 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-15.svg" alt="15" width="52" height="28" /> 条 |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> [**真实世界**](#real-world) | AgiBot Genie02 | 20 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-2000.svg" alt="2,000" width="76" height="28" /> 条 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-20.svg" alt="20" width="52" height="28" /> 条 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> [**仿真世界**](#simulation) | ARX | 15 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-1500.svg" alt="1,500" width="76" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> 条 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-simulation-15.svg" alt="15" width="52" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> 条 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> [**真实世界**](#real-world) | AgiBot Genie02 | 20 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-2000.svg" alt="2,000" width="76" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> 条 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/count-real-world-20.svg" alt="20" width="52" height="28" style="display:inline-block;margin:0;vertical-align:middle" /> 条 |
 
 | 使用设置 | 示例数量 | 用途 |
 | --- | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="16" height="16" /> **Full-Shot** | 每任务 100 条 | 使用完整专家示例进行训练或微调 |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="16" height="16" /> **One-Shot** | 每任务 1 条 | 以单条任务示例进行参考条件输入或任务适配 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **Full-Shot** | 每任务 100 条 | 使用完整专家示例进行训练或微调 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **One-Shot** | 每任务 1 条 | 以单条任务示例进行参考条件输入或任务适配 |
 
 两个 One-Shot 子集合计 **35 条轨迹、23,747 个时间步、105 个相机视频**。本数据集提供标准域内专家示��；价值或进度监督由训练方法构造。
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="24" height="24" /> 2. 仿真世界 <a name="simulation"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 2. 仿真世界 <a name="simulation"></a>
 
 ### 2.1 总览
 
@@ -42,11 +42,11 @@ Full-Shot 与 One-Shot 中，每条 **episode** 都是一条完整的任务示
 
 | 内容 | 表示 |
 | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" /> 机器人状态 | `observation.state`：`float32[14]` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" /> 机器人动作 | `action`：`float32[14]` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> RGB 观测 | 高位、左腕、右腕三路视角，均为 **640 × 480** |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" /> 指令与标识 | 任务指令、`task_index`、`episode_index`、`frame_index` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 时间 | `timestamp`，单位为秒；**25 FPS** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 机器人状态 | `observation.state`：`float32[14]` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 机器人动作 | `action`：`float32[14]` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> RGB 观测 | 高位、左腕、右腕三路视角，均为 **640 × 480** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 指令与标识 | 任务指令、`task_index`、`episode_index`、`frame_index` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间 | `timestamp`，单位为秒；**25 FPS** |
 
 <details>
 <summary><strong>仿真字段参考：状态、动作与相机</strong></summary>
@@ -87,27 +87,27 @@ Full-Shot 与 One-Shot 中，每条 **episode** 都是一条完整的任务示
 
 </details>
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="20" height="20" /> 2.2 Full-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-simulation.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 2.2 Full-Shot
 
 **15 个任务 × 100 条 = 1,500 条轨迹。**
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> 训练划分 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 训练划分 |
 | ---: | --- | ---: | ---: | --- |
 | **789,612** | **4,500 个 · AV1 / H.264** | **≈45.47 GB** | ≈8.77 小时 | `train: 0:1500` |
 
 [**浏览文件 →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/simulation/full-shot) · [下载与读取](#download-and-reading)
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="20" height="20" /> 2.3 One-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-simulation.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 2.3 One-Shot
 
 **15 个任务 × 1 条 = 15 条轨迹。** 与 Full-Shot 使用同一组任务定义。
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> 训练划分 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 训练划分 |
 | ---: | --- | ---: | ---: | --- |
 | **8,080** | **45 个 · AV1** | **≈0.50 GB** | ≈5.39 分钟 | `train: 0:15` |
 
 [**浏览文件 →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/simulation/one-shot) · [任务—轨迹—行范围映射](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/blob/main/data/simulation/one-shot/provenance/episode_mapping.json) · [下载与读取](#download-and-reading)
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="24" height="24" /> 3. 真实世界 <a name="real-world"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 3. 真实世界 <a name="real-world"></a>
 
 ### 3.1 总览
 
@@ -119,11 +119,11 @@ Full-Shot 与 One-Shot 中，每条 **episode** 都是一条完整的任务示
 
 | 内容 | 表示 |
 | --- | --- |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" /> 机器人状态 | `observation.state`：`float32[290]`，包含关节与末端状态及辅助字段 |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" /> 机器人动作 | `action`：`float32[52]`，包含关节目标、末端目标位姿及其他控制字段 |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> RGB 观测 | 头部 **640 × 400**；左腕与右腕各 **640 × 528** |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" /> 指令与标识 | 任务指令、`task_index`、`episode_index`、`frame_index` |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 时间 | `timestamp`，单位为秒；**30 FPS** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/state.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 机器人状态 | `observation.state`：`float32[290]`，包含关节与末端状态及辅助字段 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/action.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 机器人动作 | `action`：`float32[52]`，包含关节目标、末端目标位姿及其他控制字段 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> RGB 观测 | 头部 **640 × 400**；左腕与右腕各 **640 × 528** |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/instruction.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 指令与标识 | 任务指令、`task_index`、`episode_index`、`frame_index` |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间 | `timestamp`，单位为秒；**30 FPS** |
 
 相机字段为 `observation.images.cam_high`、`observation.images.cam_left_wrist` 和 `observation.images.cam_right_wrist`。读取后的图像采用 **CHW** 顺序。
 
@@ -203,27 +203,27 @@ Full-Shot 与 One-Shot 中，每条 **episode** 都是一条完整的任务示
 
 </details>
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-real-world.svg" alt="" width="20" height="20" /> 3.2 Full-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/full-shot-real-world.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 3.2 Full-Shot
 
 **20 个任务 × 100 条 = 2,000 条轨迹。**
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> 训练划分 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 训练划分 |
 | ---: | --- | ---: | ---: | --- |
 | **1,530,987** | **6,000 个 · HEVC** | **≈23.23 GB** | ≈14.18 小时 | `train: 0:2000` |
 
 [**浏览文件 →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/real-world/full-shot) · [下载与读取](#download-and-reading)
 
-### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-real-world.svg" alt="" width="20" height="20" /> 3.3 One-Shot
+### <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/one-shot-real-world.svg" alt="" width="20" height="20" style="display:inline-block;margin:0;vertical-align:middle" /> 3.3 One-Shot
 
 **20 个任务 × 1 条 = 20 条轨迹。** 与 Full-Shot 使用同一组任务定义。
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" /> 训练划分 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/video.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 视频 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 大小 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/time.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 总时长 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/split.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 训练划分 |
 | ---: | --- | ---: | ---: | --- |
 | **15,667** | **60 个 · HEVC** | **≈0.23 GB** | ≈8.70 分钟 | `train: 0:20` |
 
 [**浏览文件 →**](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/tree/main/data/real-world/one-shot) · [任务—轨迹—行范围映射](https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/blob/main/data/real-world/one-shot/provenance/episode_mapping.json) · [下载与读取](#download-and-reading)
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="24" height="24" /> 4. 仓库目录 <a name="repository-layout"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 4. 仓库目录 <a name="repository-layout"></a>
 
 ```text
 README.md                         英文数据集卡片
@@ -260,18 +260,18 @@ data/
 
 </details>
 
-## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="24" height="24" /> 5. 下载与读取 <a name="download-and-reading"></a>
+## <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="24" height="24" style="display:inline-block;margin:0;vertical-align:middle" /> 5. 下载与读取 <a name="download-and-reading"></a>
 
 ### 5.1 选择子集
 
 只需选择一个 `prefix`；下载范围限定在该子集内。大小采用十进制 GB。
 
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="16" height="16" /> 子集 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="16" height="16" /> `prefix` | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" /> 轨迹 / 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" /> 大小 |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/repository.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 子集 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/download.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> `prefix` | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/frames.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 轨迹 / 时间步 | <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/storage.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> 大小 |
 | --- | --- | ---: | ---: |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> **仿真 Full-Shot** | `data/simulation/full-shot` | 1,500 / 789,612 | ≈45.47 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" /> **仿真 One-Shot** | `data/simulation/one-shot` | 15 / 8,080 | ≈0.50 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> **真机 Full-Shot** | `data/real-world/full-shot` | 2,000 / 1,530,987 | ≈23.23 GB |
-| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" /> **真机 One-Shot** | `data/real-world/one-shot` | 20 / 15,667 | ≈0.23 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **仿真 Full-Shot** | `data/simulation/full-shot` | 1,500 / 789,612 | ≈45.47 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/simulation.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **仿真 One-Shot** | `data/simulation/one-shot` | 15 / 8,080 | ≈0.50 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **真机 Full-Shot** | `data/real-world/full-shot` | 2,000 / 1,530,987 | ≈23.23 GB |
+| <img src="https://huggingface.co/datasets/wzlllong/value-model-benchmark-training-data/resolve/main/assets/card-design-v1/real-world.svg" alt="" width="16" height="16" style="display:inline-block;margin:0;vertical-align:middle" /> **真机 One-Shot** | `data/real-world/one-shot` | 20 / 15,667 | ≈0.23 GB |
 
 ### 5.2 下载并读取