add dataset card README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,91 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
| 3 |
+
viewer: false
|
| 4 |
+
pretty_name: ModerationPrompts
|
| 5 |
+
tags:
|
| 6 |
+
- content-moderation
|
| 7 |
+
- prompts
|
| 8 |
+
- llm
|
| 9 |
+
- safety
|
| 10 |
+
- bluesky
|
| 11 |
+
- social-media
|
| 12 |
---
|
| 13 |
+
|
| 14 |
+
# ModerationPrompts
|
| 15 |
+
|
| 16 |
+
System and user prompts used for LLM-based content moderation.
|
| 17 |
+
Each prompt instructs a vision-language model to classify a social-media post (text + image(s))
|
| 18 |
+
against a safety policy and return a structured JSON decision.
|
| 19 |
+
|
| 20 |
+
---
|
| 21 |
+
|
| 22 |
+
## Repository Structure
|
| 23 |
+
|
| 24 |
+
```
|
| 25 |
+
prompts/
|
| 26 |
+
bluesky_policy/ ← prompts grounded in Bluesky community guidelines
|
| 27 |
+
community_guidelines.md policy document (included verbatim in some prompts)
|
| 28 |
+
prompt_labels.md
|
| 29 |
+
prompt_labels_details.md
|
| 30 |
+
prompt_labels_rationale.md
|
| 31 |
+
prompt_labels_rationale_details.md
|
| 32 |
+
meta_policy/ ← prompts grounded in Meta content policy
|
| 33 |
+
all_category_w_explanation.md all harm categories with detailed explanations
|
| 34 |
+
```
|
| 35 |
+
|
| 36 |
+
---
|
| 37 |
+
|
| 38 |
+
## Bluesky-Policy Prompts
|
| 39 |
+
|
| 40 |
+
| File | Includes community guidelines | Label detail level |
|
| 41 |
+
|---|:---:|:---:|
|
| 42 |
+
| `community_guidelines.md` | — (guidelines doc only) | — |
|
| 43 |
+
| `prompt_labels.md` | No | Basic |
|
| 44 |
+
| `prompt_labels_details.md` | No | Detailed (with sub-rules) |
|
| 45 |
+
| `prompt_labels_rationale.md` | Yes | Basic |
|
| 46 |
+
| `prompt_labels_rationale_details.md` | Yes | Detailed (with sub-rules) |
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
---
|
| 50 |
+
|
| 51 |
+
## Meta-Policy Prompts
|
| 52 |
+
|
| 53 |
+
| File | Policy source | All harm categories | Includes explanations | Requests rationale |
|
| 54 |
+
|---|:---:|:---:|:---:|:---:|
|
| 55 |
+
| `all_category_w_explanation.md` | Meta community standards | Yes | Yes | Yes |
|
| 56 |
+
|
| 57 |
+
**`all_category_w_explanation.md`** — full prompt grounded in Meta content
|
| 58 |
+
standards. Covers all harm categories with per-category explanations.
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
|
| 62 |
+
## Prompt Format
|
| 63 |
+
|
| 64 |
+
Each `.md` file contains clearly delimited sections:
|
| 65 |
+
|
| 66 |
+
```
|
| 67 |
+
SYSTEM PROMPT:
|
| 68 |
+
<system instruction — role, task, output format constraint>
|
| 69 |
+
|
| 70 |
+
------------
|
| 71 |
+
|
| 72 |
+
USER PROMPT:
|
| 73 |
+
# **COMMUNITY GUIDELINES** ← present in _rationale and meta_policy variants
|
| 74 |
+
...
|
| 75 |
+
|
| 76 |
+
# **LABELING POLICY DETAILS**
|
| 77 |
+
## **S1**: porn
|
| 78 |
+
...
|
| 79 |
+
|
| 80 |
+
POST CONTENT:
|
| 81 |
+
<post text and image placeholder>
|
| 82 |
+
|
| 83 |
+
OUTPUT FORMAT:
|
| 84 |
+
<required JSON schema>
|
| 85 |
+
```
|
| 86 |
+
|
| 87 |
+
---
|
| 88 |
+
|
| 89 |
+
## License
|
| 90 |
+
|
| 91 |
+
CC BY-NC 4.0 — non-commercial use only.
|