File size: 2,199 Bytes
f9b150b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
# Migration to Salesforce/BehaviorBench

The canonical release destination is:

```text
https://huggingface.co/datasets/Salesforce/BehaviorBench
```

Use the provided export script rather than copying the repository manually. It exports only the Polymarket BehaviorBench artifact and excludes `manifold/`, Git internals, caches, and unrelated local files.

## 1. Download the temporary repository with LFS objects

```bash
git lfs install
git clone https://huggingface.co/datasets/lyang84/BehaviorBench-camera-ready
cd BehaviorBench-camera-ready
git lfs pull
git lfs fsck
```

Do not use a source-code ZIP: large Parquet files are stored with Git LFS.

## 2. Build the clean Salesforce release directory

From the downloaded repository:

```bash
python scripts/prepare_salesforce_release.py --output ../BehaviorBench-release
cd ../BehaviorBench-release
python scripts/validate_artifact.py
```

The exporter uses `croissant.json` as the authoritative data-file manifest and also copies release documentation, compressed JSONL benchmark files, metadata, wallet lists, evaluation files, and scripts.

## 3. Upload without changing paths

Create the destination dataset repo, then:

```bash
cd BehaviorBench-release
git init
git lfs install
git lfs track "*.parquet" "*.jsonl.gz"
git add .
git commit -m "Release BehaviorBench artifact"
git branch -M main
git remote add origin https://huggingface.co/datasets/Salesforce/BehaviorBench
git push -u origin main
```

Do not rename `dataset/`, `benchmark/`, or files listed in `croissant.json`.

## 4. Validate the final hosted artifact

After the upload completes:

```bash
python scripts/check_hosted_artifact.py
```

This checks the final Salesforce dataset URL and every Croissant `contentUrl`. Then run the NeurIPS online Croissant validator and upload the validated `croissant.json` to OpenReview.

## Failure prevention

- Ensure Git LFS is installed before download and upload.
- Confirm Parquet files are real binaries, not small LFS pointer files.
- Preserve case-sensitive paths and the `main` branch.
- Do not include `manifold/` in the Salesforce repo.
- Do not edit `croissant.json` URLs unless the final organization/repository name changes.