Update README to add further results
Browse files
README.md
CHANGED
|
@@ -1,8 +1,19 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: other
|
| 3 |
-
license_name: govtech-singapore
|
| 4 |
-
license_link: LICENSE
|
| 5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
# Off-Topic Classification Model
|
| 8 |
|
|
@@ -16,10 +27,18 @@ This model leverages a fine-tuned **Cross Encoder STSB Roberta Base** to perform
|
|
| 16 |
|
| 17 |
## Performance
|
| 18 |
|
|
|
|
|
|
|
| 19 |
| Approach | Model | ROC-AUC | F1 | Precision | Recall |
|
| 20 |
|---------------------------------------|--------------------------------|---------|------|-----------|--------|
|
| 21 |
-
| Fine-tuned
|
|
|
|
| 22 |
| Pre-trained cross-encoder | stsb-roberta-base | 0.73 | 0.68 | 0.53 | 0.93 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
|
| 24 |
## Usage
|
| 25 |
1. Clone this repository and install the required dependencies:
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: govtech-singapore
|
| 4 |
+
license_link: LICENSE
|
| 5 |
+
datasets:
|
| 6 |
+
- gabrielchua/off-topic
|
| 7 |
+
language:
|
| 8 |
+
- en
|
| 9 |
+
metrics:
|
| 10 |
+
- roc_auc
|
| 11 |
+
- f1
|
| 12 |
+
- precision
|
| 13 |
+
- recall
|
| 14 |
+
base_model:
|
| 15 |
+
- cross-encoder/stsb-roberta-base
|
| 16 |
+
---
|
| 17 |
|
| 18 |
# Off-Topic Classification Model
|
| 19 |
|
|
|
|
| 27 |
|
| 28 |
## Performance
|
| 29 |
|
| 30 |
+
We evaluated our fine-tuned models on synthetic data modelling system and user prompt pairs reflecting real world enterprise use cases of LLMs. The dataset is available [here](https://huggingface.co/datasets/gabrielchua/off-topic).
|
| 31 |
+
|
| 32 |
| Approach | Model | ROC-AUC | F1 | Precision | Recall |
|
| 33 |
|---------------------------------------|--------------------------------|---------|------|-----------|--------|
|
| 34 |
+
| 👉 [Fine-tuned bi-encoder classifier](https://huggingface.co/govtech/jina-embeddings-v2-small-en-off-topic) | jina-embeddings-v2-small-en | 0.99 | 0.97 | 0.99 | 0.95 |
|
| 35 |
+
| [Fine-tuned cross-encoder classifier](https://huggingface.co/govtech/stsb-roberta-base-off-topic) | stsb-roberta-base | 0.99 | 0.99 | 0.99 | 0.99 |
|
| 36 |
| Pre-trained cross-encoder | stsb-roberta-base | 0.73 | 0.68 | 0.53 | 0.93 |
|
| 37 |
+
| Prompt Engineering | GPT 4o (2024-08-06) | - | 0.95 | 0.94 | 0.97 |
|
| 38 |
+
| Prompt Engineering | GPT 4o Mini (2024-07-18) | - | 0.91 | 0.85 | 0.91 |
|
| 39 |
+
| Zero-shot Classification | GPT 4o Mini (2024-07-18) | 0.99 | 0.97 | 0.95 | 0.99 |
|
| 40 |
+
|
| 41 |
+
Further evaluation results on additional synthetic and external datasets (e.g.,`JailbreakBench`, `HarmBench`, `TrustLLM`) are available in our [technical report](https://arxiv.org/abs/2411.12946).
|
| 42 |
|
| 43 |
## Usage
|
| 44 |
1. Clone this repository and install the required dependencies:
|