Link the blog write-up at the top of the dataset card
Browse filesAdds a callout under the title linking to the WGO-Bench write-up (with UTM tags so visits from the card are attributable) and the Refiner repo.
README.md
CHANGED
|
@@ -27,6 +27,8 @@ configs:
|
|
| 27 |
|
| 28 |
# WGO-Bench: What's Going On Benchmark
|
| 29 |
|
|
|
|
|
|
|
| 30 |
WGO-Bench is a small, manually annotated benchmark for evaluating how well vision-language models can turn robot and egocentric manipulation videos into timestamped subtask annotations.
|
| 31 |
|
| 32 |
Each row contains one video episode, a high-level task instruction, and gold subtask segments with start time, end time, and a concise action label. The benchmark is designed for two related tasks:
|
|
|
|
| 27 |
|
| 28 |
# WGO-Bench: What's Going On Benchmark
|
| 29 |
|
| 30 |
+
> **Read the write-up:** [Segmenting Robot Video into Actionable Subtasks](https://macrodata.co/blog/annotating-robot-video-subtasks?utm_source=huggingface&utm_medium=dataset_card&utm_campaign=wgo-bench) covers how WGO-Bench was built, the 60+ annotation pipeline experiments we ran on it, and what worked. Code: [macrodata-labs/refiner](https://github.com/macrodata-labs/refiner).
|
| 31 |
+
|
| 32 |
WGO-Bench is a small, manually annotated benchmark for evaluating how well vision-language models can turn robot and egocentric manipulation videos into timestamped subtask annotations.
|
| 33 |
|
| 34 |
Each row contains one video episode, a high-level task instruction, and gold subtask segments with start time, end time, and a concise action label. The benchmark is designed for two related tasks:
|