Simplify dataset documentation and update framework compatibility
Browse files- CITATION.cff +14 -29
- DATA_DICTIONARY.md +1 -1
- PACKAGE_MANIFEST.json +1 -1
- README.md +8 -17
- SHA256SUMS +4 -4
CITATION.cff
CHANGED
|
@@ -1,33 +1,18 @@
|
|
| 1 |
cff-version: 1.2.0
|
| 2 |
-
message:
|
| 3 |
type: dataset
|
| 4 |
-
title:
|
| 5 |
-
version:
|
| 6 |
-
url:
|
| 7 |
license: CC-BY-4.0
|
| 8 |
authors:
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
preferred-citation:
|
| 20 |
-
type: article
|
| 21 |
-
title: "CONFIRM: Governing Scientific Reporting for Trustworthy Agentic Neuroimaging"
|
| 22 |
-
authors:
|
| 23 |
-
- family-names: Wei
|
| 24 |
-
given-names: Yuxiang
|
| 25 |
-
- family-names: Zhu
|
| 26 |
-
given-names: Lijing
|
| 27 |
-
- family-names: Lan
|
| 28 |
-
given-names: Qizhen
|
| 29 |
-
- family-names: Zhang
|
| 30 |
-
given-names: Yanteng
|
| 31 |
-
- family-names: Calhoun
|
| 32 |
-
given-names: Vince D.
|
| 33 |
-
url: "https://openreview.net/forum?id=KYE1DzsxWc"
|
|
|
|
| 1 |
cff-version: 1.2.0
|
| 2 |
+
message: Citation metadata for NeuroClaimBench.
|
| 3 |
type: dataset
|
| 4 |
+
title: NeuroClaimBench
|
| 5 |
+
version: '2.1'
|
| 6 |
+
url: https://huggingface.co/datasets/stanjsx/NeuroClaimBench
|
| 7 |
license: CC-BY-4.0
|
| 8 |
authors:
|
| 9 |
+
- family-names: Wei
|
| 10 |
+
given-names: Yuxiang
|
| 11 |
+
- family-names: Zhu
|
| 12 |
+
given-names: Lijing
|
| 13 |
+
- family-names: Lan
|
| 14 |
+
given-names: Qizhen
|
| 15 |
+
- family-names: Zhang
|
| 16 |
+
given-names: Yanteng
|
| 17 |
+
- family-names: Calhoun
|
| 18 |
+
given-names: Vince D.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
DATA_DICTIONARY.md
CHANGED
|
@@ -101,6 +101,6 @@ cohort files.
|
|
| 101 |
| `benchmark_manifest.json` | Record filenames, hashes, schema version, counts, and source identities |
|
| 102 |
| `RELEASE_MANIFEST.json` | Benchmark identity, record counts, and distributed file inventory |
|
| 103 |
| `PACKAGE_MANIFEST.json` | Version, license, framework dependency, and core record hashes |
|
| 104 |
-
| `CITATION.cff` |
|
| 105 |
| `LICENSE` | CC BY 4.0 license text |
|
| 106 |
| `SHA256SUMS` | Checksums of every distributed file except this checksum file |
|
|
|
|
| 101 |
| `benchmark_manifest.json` | Record filenames, hashes, schema version, counts, and source identities |
|
| 102 |
| `RELEASE_MANIFEST.json` | Benchmark identity, record counts, and distributed file inventory |
|
| 103 |
| `PACKAGE_MANIFEST.json` | Version, license, framework dependency, and core record hashes |
|
| 104 |
+
| `CITATION.cff` | Dataset authors and version |
|
| 105 |
| `LICENSE` | CC BY 4.0 license text |
|
| 106 |
| `SHA256SUMS` | Checksums of every distributed file except this checksum file |
|
PACKAGE_MANIFEST.json
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
"version": "2.1",
|
| 4 |
"schema_version": 2,
|
| 5 |
"license": "CC-BY-4.0",
|
| 6 |
-
"framework_version": "0.1.
|
| 7 |
"frozen_record_sha256": {
|
| 8 |
"cases": "57605904c124ecadbc9105d9b55e3bba58ee94aab0e68d6a5f93e587f6359c41",
|
| 9 |
"references": "90f6cf8595784c84afb80020b071ab6521931c6382365cd60da7f6452ec3e137",
|
|
|
|
| 3 |
"version": "2.1",
|
| 4 |
"schema_version": 2,
|
| 5 |
"license": "CC-BY-4.0",
|
| 6 |
+
"framework_version": "0.1.1",
|
| 7 |
"frozen_record_sha256": {
|
| 8 |
"cases": "57605904c124ecadbc9105d9b55e3bba58ee94aab0e68d6a5f93e587f6359c41",
|
| 9 |
"references": "90f6cf8595784c84afb80020b071ab6521931c6382365cd60da7f6452ec3e137",
|
README.md
CHANGED
|
@@ -33,8 +33,7 @@ configs:
|
|
| 33 |
# NeuroClaimBench v2.1
|
| 34 |
|
| 35 |
NeuroClaimBench evaluates whether a system supports or withholds a
|
| 36 |
-
neuroimaging claim.
|
| 37 |
-
[CONFIRM: Governing Scientific Reporting for Trustworthy Agentic Neuroimaging](https://openreview.net/forum?id=KYE1DzsxWc).
|
| 38 |
|
| 39 |
The package contains 489 executable tasks: 279 scientific claims, 150
|
| 40 |
constructed stress controls, and 60 external-transfer claims and controls.
|
|
@@ -58,7 +57,7 @@ Load an individual record collection:
|
|
| 58 |
```python
|
| 59 |
from datasets import load_dataset
|
| 60 |
|
| 61 |
-
tasks = load_dataset("stanjsx/NeuroClaimBench", "tasks", split="test", revision="v2.1")
|
| 62 |
print(tasks[0]["task_id"])
|
| 63 |
print(tasks[0]["contract"])
|
| 64 |
```
|
|
@@ -76,7 +75,7 @@ from huggingface_hub import snapshot_download
|
|
| 76 |
snapshot_download(
|
| 77 |
repo_id="stanjsx/NeuroClaimBench",
|
| 78 |
repo_type="dataset",
|
| 79 |
-
revision="v2.1",
|
| 80 |
local_dir="neuroclaimbench-v2.1",
|
| 81 |
)
|
| 82 |
```
|
|
@@ -84,19 +83,19 @@ snapshot_download(
|
|
| 84 |
## Install the tools
|
| 85 |
|
| 86 |
Use Python 3.10 or newer. CONFIRM supplies the `neuroclaimbench` command.
|
| 87 |
-
Download `confirm-0.1.
|
| 88 |
-
[CONFIRM v0.1.
|
| 89 |
into the directory containing the extracted `neuroclaimbench-v2.1` folder.
|
| 90 |
Run from that parent directory:
|
| 91 |
|
| 92 |
```bash
|
| 93 |
python -m venv .venv
|
| 94 |
source .venv/bin/activate
|
| 95 |
-
python -m pip install ./confirm-0.1.
|
| 96 |
```
|
| 97 |
|
| 98 |
-
|
| 99 |
-
extract `confirm-0.1.
|
| 100 |
from its parent directory. An existing CONFIRM installation provides the
|
| 101 |
same benchmark commands.
|
| 102 |
|
|
@@ -121,15 +120,7 @@ no cohort data or model calls. Choose a new output filename for each export.
|
|
| 121 |
- [Python loading, metrics, and evaluation](USAGE.md)
|
| 122 |
- [File and field definitions](DATA_DICTIONARY.md)
|
| 123 |
- [Cohort data and exact-input requirements](DATA_ACCESS.md)
|
| 124 |
-
- [Citation](CITATION.cff)
|
| 125 |
|
| 126 |
Executing a task requires its corresponding prepared cohort files. The
|
| 127 |
metadata package contains no participant tables, random-control assignment
|
| 128 |
tables, or PubMed abstracts.
|
| 129 |
-
|
| 130 |
-
## License
|
| 131 |
-
|
| 132 |
-
The authors' benchmark metadata and documentation are licensed under
|
| 133 |
-
[CC BY 4.0](LICENSE). Source datasets and referenced third-party works retain
|
| 134 |
-
their own licenses and access terms. Cite the paper and this benchmark when
|
| 135 |
-
using the release.
|
|
|
|
| 33 |
# NeuroClaimBench v2.1
|
| 34 |
|
| 35 |
NeuroClaimBench evaluates whether a system supports or withholds a
|
| 36 |
+
neuroimaging claim.
|
|
|
|
| 37 |
|
| 38 |
The package contains 489 executable tasks: 279 scientific claims, 150
|
| 39 |
constructed stress controls, and 60 external-transfer claims and controls.
|
|
|
|
| 57 |
```python
|
| 58 |
from datasets import load_dataset
|
| 59 |
|
| 60 |
+
tasks = load_dataset("stanjsx/NeuroClaimBench", "tasks", split="test", revision="v2.1.1")
|
| 61 |
print(tasks[0]["task_id"])
|
| 62 |
print(tasks[0]["contract"])
|
| 63 |
```
|
|
|
|
| 75 |
snapshot_download(
|
| 76 |
repo_id="stanjsx/NeuroClaimBench",
|
| 77 |
repo_type="dataset",
|
| 78 |
+
revision="v2.1.1",
|
| 79 |
local_dir="neuroclaimbench-v2.1",
|
| 80 |
)
|
| 81 |
```
|
|
|
|
| 83 |
## Install the tools
|
| 84 |
|
| 85 |
Use Python 3.10 or newer. CONFIRM supplies the `neuroclaimbench` command.
|
| 86 |
+
Download `confirm-0.1.1-py3-none-any.whl` from the
|
| 87 |
+
[CONFIRM v0.1.1 release](https://github.com/yuxiangwei0808/CONFIRM/releases/tag/v0.1.1)
|
| 88 |
into the directory containing the extracted `neuroclaimbench-v2.1` folder.
|
| 89 |
Run from that parent directory:
|
| 90 |
|
| 91 |
```bash
|
| 92 |
python -m venv .venv
|
| 93 |
source .venv/bin/activate
|
| 94 |
+
python -m pip install ./confirm-0.1.1-py3-none-any.whl
|
| 95 |
```
|
| 96 |
|
| 97 |
+
To install from source,
|
| 98 |
+
extract `confirm-0.1.1.zip` and run `python -m pip install ./confirm-0.1.1`
|
| 99 |
from its parent directory. An existing CONFIRM installation provides the
|
| 100 |
same benchmark commands.
|
| 101 |
|
|
|
|
| 120 |
- [Python loading, metrics, and evaluation](USAGE.md)
|
| 121 |
- [File and field definitions](DATA_DICTIONARY.md)
|
| 122 |
- [Cohort data and exact-input requirements](DATA_ACCESS.md)
|
|
|
|
| 123 |
|
| 124 |
Executing a task requires its corresponding prepared cohort files. The
|
| 125 |
metadata package contains no participant tables, random-control assignment
|
| 126 |
tables, or PubMed abstracts.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
SHA256SUMS
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
-
|
| 2 |
b03098ab30c8723e556d3e6383b0df944b6c39cee8a543fa90d38986d131cbcc DATA_ACCESS.md
|
| 3 |
-
|
| 4 |
d557539df68e771cc1eedcc91d13f70fca930e508d11eedcafa4b15db49e3744 LICENSE
|
| 5 |
-
|
| 6 |
-
|
| 7 |
119ae933b7959d1907fe0e97be724686d47f660e85feecd2a9b1f4b3c8286f1d RELEASE_MANIFEST.json
|
| 8 |
c60a681c5e3e6d019deb32b220b98edda8cbde497c902726a9e21f69f8b3998d USAGE.md
|
| 9 |
ac738bbcd45c44dc572202509af06445f81dafeb9a848df0142c1981489e58bd alignment_policy.json
|
|
|
|
| 1 |
+
6b7b05a71921d5525e7b8b3e370a3963cc428368143bb6042dbd2e007af01d0d CITATION.cff
|
| 2 |
b03098ab30c8723e556d3e6383b0df944b6c39cee8a543fa90d38986d131cbcc DATA_ACCESS.md
|
| 3 |
+
941390cf882bd4efbc7667f6a5e9f1edddcf29cff22e9a7ed99e3bce9ea97a22 DATA_DICTIONARY.md
|
| 4 |
d557539df68e771cc1eedcc91d13f70fca930e508d11eedcafa4b15db49e3744 LICENSE
|
| 5 |
+
7c5ca85188426a993e5715175f9fc044a4c49dc772f091f27386fe46f2995b93 PACKAGE_MANIFEST.json
|
| 6 |
+
b6526526478d30c821573f0e089a803233db98278c777d3308e3a2b1f58a6976 README.md
|
| 7 |
119ae933b7959d1907fe0e97be724686d47f660e85feecd2a9b1f4b3c8286f1d RELEASE_MANIFEST.json
|
| 8 |
c60a681c5e3e6d019deb32b220b98edda8cbde497c902726a9e21f69f8b3998d USAGE.md
|
| 9 |
ac738bbcd45c44dc572202509af06445f81dafeb9a848df0142c1981489e58bd alignment_policy.json
|