Update README.md
Browse files
README.md
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
-
pretty_name:
|
| 6 |
tags:
|
| 7 |
- satellite
|
| 8 |
- communications
|
|
@@ -15,10 +15,10 @@ size_categories:
|
|
| 15 |
|
| 16 |
## Dataset Summary
|
| 17 |
|
| 18 |
-
**SATCOM-QA** is a **human-generated evaluation dataset** designed for testing LLM reasoning in
|
| 19 |
-
It contains
|
| 20 |
|
| 21 |
-
The dataset is
|
| 22 |
|
| 23 |
---
|
| 24 |
|
|
@@ -35,7 +35,7 @@ The dataset is intended **purely for evaluation**.
|
|
| 35 |
* Multi-layer NTNs (satellites, HAPS, UAVs)
|
| 36 |
* 6G integration
|
| 37 |
* Localization with NTNs
|
| 38 |
-
*
|
| 39 |
* Includes **reference links** to the answers provided
|
| 40 |
* No synthetic content, no model-generated text
|
| 41 |
|
|
@@ -51,7 +51,6 @@ The dataset is intended **purely for evaluation**.
|
|
| 51 |
* Telecommunications academia/industry
|
| 52 |
* Evaluating mathematical reasoning in real-world SatCom tasks
|
| 53 |
|
| 54 |
-
Not intended for training or commercial deployment.
|
| 55 |
|
| 56 |
---
|
| 57 |
|
|
@@ -67,13 +66,6 @@ Not intended for training or commercial deployment.
|
|
| 67 |
|
| 68 |
---
|
| 69 |
|
| 70 |
-
## Dataset Size
|
| 71 |
-
|
| 72 |
-
* around 1,000 total samples
|
| 73 |
-
* around 300 math-related SatCom questions
|
| 74 |
-
|
| 75 |
-
---
|
| 76 |
-
|
| 77 |
## Source Material
|
| 78 |
|
| 79 |
Questions are based on **public academic literature**, including (examples):
|
|
@@ -89,5 +81,4 @@ from datasets import load_dataset
|
|
| 89 |
ds = load_dataset("esa-sceva/satcom-qa")
|
| 90 |
test_ds = ds["train"]
|
| 91 |
print(test_ds[0]["question"])
|
| 92 |
-
print(test_ds[0]["answer"])
|
| 93 |
-
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
+
pretty_name: satcom_qa
|
| 6 |
tags:
|
| 7 |
- satellite
|
| 8 |
- communications
|
|
|
|
| 15 |
|
| 16 |
## Dataset Summary
|
| 17 |
|
| 18 |
+
**SATCOM-QA** is a **human-generated evaluation dataset** designed for testing LLM reasoning in Satellite Communications (SatCom).
|
| 19 |
+
It contains around 1,000 open-ended questions**, of which about 30% involve mathematical or quantitative SatCom reasoning (e.g., link budget calculations, path loss, beamforming, SNR, orbital geometry).
|
| 20 |
|
| 21 |
+
The dataset is designed for evaluation purpose.
|
| 22 |
|
| 23 |
---
|
| 24 |
|
|
|
|
| 35 |
* Multi-layer NTNs (satellites, HAPS, UAVs)
|
| 36 |
* 6G integration
|
| 37 |
* Localization with NTNs
|
| 38 |
+
* 30% math-focused problems
|
| 39 |
* Includes **reference links** to the answers provided
|
| 40 |
* No synthetic content, no model-generated text
|
| 41 |
|
|
|
|
| 51 |
* Telecommunications academia/industry
|
| 52 |
* Evaluating mathematical reasoning in real-world SatCom tasks
|
| 53 |
|
|
|
|
| 54 |
|
| 55 |
---
|
| 56 |
|
|
|
|
| 66 |
|
| 67 |
---
|
| 68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
## Source Material
|
| 70 |
|
| 71 |
Questions are based on **public academic literature**, including (examples):
|
|
|
|
| 81 |
ds = load_dataset("esa-sceva/satcom-qa")
|
| 82 |
test_ds = ds["train"]
|
| 83 |
print(test_ds[0]["question"])
|
| 84 |
+
print(test_ds[0]["answer"])
|
|
|