Upload README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
These two classifier models are fine-tuned to flag possible manipulation in messages, having been trained on synthetic interpersonal relationship data.
|
| 2 |
|
| 3 |
The smaller model is based on microsoft/xtremedistil-l6-h256-uncased and has 12.75M total parameters. The larger uses microsoft/deberta-v3-xsmall and is at 70.83M total parameters. Both models achieve +99% F1 score on the held out test split.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
license: mit
|
| 5 |
+
tags:
|
| 6 |
+
- text-classification
|
| 7 |
+
- manipulation-detection
|
| 8 |
+
- pytorch
|
| 9 |
+
- transformers
|
| 10 |
+
- interpersonal-relationships
|
| 11 |
+
library_name: transformers
|
| 12 |
+
pipeline_tag: text-classification
|
| 13 |
+
metrics:
|
| 14 |
+
- f1
|
| 15 |
+
- accuracy
|
| 16 |
+
- precision
|
| 17 |
+
- recall
|
| 18 |
+
model-index:
|
| 19 |
+
- name: manipulation-detector-xtremedistil
|
| 20 |
+
results:
|
| 21 |
+
- task:
|
| 22 |
+
type: text-classification
|
| 23 |
+
name: Manipulation Detection
|
| 24 |
+
dataset:
|
| 25 |
+
name: synthetic-interpersonal-data
|
| 26 |
+
type: text-classification
|
| 27 |
+
metrics:
|
| 28 |
+
- type: f1
|
| 29 |
+
value: 0.99
|
| 30 |
+
- type: accuracy
|
| 31 |
+
value: 0.99
|
| 32 |
+
- name: manipulation-detector-deberta
|
| 33 |
+
results:
|
| 34 |
+
- task:
|
| 35 |
+
type: text-classification
|
| 36 |
+
name: Manipulation Detection
|
| 37 |
+
dataset:
|
| 38 |
+
name: synthetic-interpersonal-data
|
| 39 |
+
type: text-classification
|
| 40 |
+
metrics:
|
| 41 |
+
- type: f1
|
| 42 |
+
value: 0.99
|
| 43 |
+
- type: accuracy
|
| 44 |
+
value: 0.99
|
| 45 |
+
---
|
| 46 |
+
|
| 47 |
These two classifier models are fine-tuned to flag possible manipulation in messages, having been trained on synthetic interpersonal relationship data.
|
| 48 |
|
| 49 |
The smaller model is based on microsoft/xtremedistil-l6-h256-uncased and has 12.75M total parameters. The larger uses microsoft/deberta-v3-xsmall and is at 70.83M total parameters. Both models achieve +99% F1 score on the held out test split.
|