Add Dataset Viewer config and flat episode index
Browse files- README.md +40 -1
- data/embodied_memory_bench.jsonl +0 -0
README.md
CHANGED
|
@@ -7,6 +7,45 @@ size_categories:
|
|
| 7 |
- 1K<n<10K
|
| 8 |
task_categories:
|
| 9 |
- robotics
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
tags:
|
| 11 |
- embodied-ai
|
| 12 |
- embodied-memory
|
|
@@ -51,7 +90,7 @@ tags:
|
|
| 51 |
| **Experience Generalization** | A pattern learned from corrections and applied to a new object or scene | **203** |
|
| 52 |
| **Total** | Four complementary memory capabilities | **2,554** |
|
| 53 |
|
| 54 |
-
Interaction Failure episodes are stored under `interaction_experience/`. The dataset contains historical RGB observations, action and feedback traces, task probes, and evaluation annotations. Benchmark membership is fixed by [the Full-2554 manifest](manifests/full2554.jsonl).
|
| 55 |
|
| 56 |
<details>
|
| 57 |
<summary><strong>How the benchmark is constructed</strong></summary>
|
|
|
|
| 7 |
- 1K<n<10K
|
| 8 |
task_categories:
|
| 9 |
- robotics
|
| 10 |
+
configs:
|
| 11 |
+
- config_name: default
|
| 12 |
+
data_files:
|
| 13 |
+
- split: all
|
| 14 |
+
path: data/embodied_memory_bench.jsonl
|
| 15 |
+
dataset_info:
|
| 16 |
+
- config_name: default
|
| 17 |
+
features:
|
| 18 |
+
- name: id
|
| 19 |
+
dtype: string
|
| 20 |
+
- name: family
|
| 21 |
+
dtype: string
|
| 22 |
+
- name: family_id
|
| 23 |
+
dtype: string
|
| 24 |
+
- name: difficulty
|
| 25 |
+
dtype: string
|
| 26 |
+
- name: episode_name
|
| 27 |
+
dtype: string
|
| 28 |
+
- name: description
|
| 29 |
+
dtype: string
|
| 30 |
+
- name: scene
|
| 31 |
+
dtype: string
|
| 32 |
+
- name: probe_id
|
| 33 |
+
dtype: string
|
| 34 |
+
- name: probe_instruction
|
| 35 |
+
dtype: string
|
| 36 |
+
- name: context_sessions
|
| 37 |
+
dtype: int32
|
| 38 |
+
- name: context_steps
|
| 39 |
+
dtype: int32
|
| 40 |
+
- name: memory_cues
|
| 41 |
+
dtype: string
|
| 42 |
+
- name: episode_path
|
| 43 |
+
dtype: string
|
| 44 |
+
- name: preview_image
|
| 45 |
+
dtype: string
|
| 46 |
+
splits:
|
| 47 |
+
- name: all
|
| 48 |
+
num_examples: 2554
|
| 49 |
tags:
|
| 50 |
- embodied-ai
|
| 51 |
- embodied-memory
|
|
|
|
| 90 |
| **Experience Generalization** | A pattern learned from corrections and applied to a new object or scene | **203** |
|
| 91 |
| **Total** | Four complementary memory capabilities | **2,554** |
|
| 92 |
|
| 93 |
+
Interaction Failure episodes are stored under `interaction_experience/`. The dataset contains historical RGB observations, action and feedback traces, task probes, and evaluation annotations. Benchmark membership is fixed by [the Full-2554 manifest](manifests/full2554.jsonl). The **Dataset Viewer** uses the compact `all` split at [`data/embodied_memory_bench.jsonl`](data/embodied_memory_bench.jsonl); open an `episode_path` to inspect the complete record and `preview_image` to view a representative observation.
|
| 94 |
|
| 95 |
<details>
|
| 96 |
<summary><strong>How the benchmark is constructed</strong></summary>
|
data/embodied_memory_bench.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|