Update README.md
Browse files
README.md
CHANGED
|
@@ -811,8 +811,9 @@ dataset_info:
|
|
| 811 |
**Text2DistBench** is a benchmark for evaluating whether large language models can infer distributional knowledge from natural language evidence.
|
| 812 |
Given metadata and a set of user comments about an entity (e.g., a movie or song), models must estimate statistics such as:
|
| 813 |
stance distribution, topic distribution, and most/second-most frequent labels.
|
| 814 |
-
The dataset is organized into two types of configurations.
|
| 815 |
|
|
|
|
|
|
|
| 816 |
**Posterior configurations** include both metadata and user comments as evidence.
|
| 817 |
These configurations differ by sample size (50 or 100 comments) and task type:
|
| 818 |
`posterior_sampled_50_estimation`,
|
|
@@ -827,8 +828,6 @@ These configurations differ by sample size (50 or 100 comments) and task type:
|
|
| 827 |
---
|
| 828 |
|
| 829 |
## Load the Dataset
|
| 830 |
-
|
| 831 |
-
Example:
|
| 832 |
```bash
|
| 833 |
from datasets import load_dataset
|
| 834 |
ds = load_dataset("frett/Text2DistBench", "posterior_sampled_50_estimation", split="test")
|
|
|
|
| 811 |
**Text2DistBench** is a benchmark for evaluating whether large language models can infer distributional knowledge from natural language evidence.
|
| 812 |
Given metadata and a set of user comments about an entity (e.g., a movie or song), models must estimate statistics such as:
|
| 813 |
stance distribution, topic distribution, and most/second-most frequent labels.
|
|
|
|
| 814 |
|
| 815 |
+
The dataset is constructed from movie and music entities released between 2025-12-01 and 2026-03-01.
|
| 816 |
+
It includes two types of configurations:
|
| 817 |
**Posterior configurations** include both metadata and user comments as evidence.
|
| 818 |
These configurations differ by sample size (50 or 100 comments) and task type:
|
| 819 |
`posterior_sampled_50_estimation`,
|
|
|
|
| 828 |
---
|
| 829 |
|
| 830 |
## Load the Dataset
|
|
|
|
|
|
|
| 831 |
```bash
|
| 832 |
from datasets import load_dataset
|
| 833 |
ds = load_dataset("frett/Text2DistBench", "posterior_sampled_50_estimation", split="test")
|