pfr commited on
Commit
fbf1471
1 Parent(s): c535f7b

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +35 -1
README.md CHANGED
@@ -6,4 +6,38 @@ inference:
6
  function_to_apply: "none"
7
  widget:
8
  - text: "I cuddled with my dog today."
9
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  function_to_apply: "none"
7
  widget:
8
  - text: "I cuddled with my dog today."
9
+ ---
10
+
11
+ # Utilitarian Deberta 01
12
+
13
+ ## Model description
14
+
15
+ This is a [Deberta model](https://huggingface.co/microsoft/deberta-v3-large) fine-tuned on for computing utility estimates of experiences, represented in first-person sentences. It was trained from human-annotated pairwise utility comparisons, from the [ETHICS dataset](https://arxiv.org/abs/2008.02275).
16
+
17
+ ## Intended use
18
+
19
+ The main use case is the computation of utility estimates of first-person text scenarios.
20
+
21
+ ## Limitations
22
+
23
+ The model was only trained on a limited number of scenarios, and only on first-person sentences. It does not have the capability of interpreting highly complex or unusual scenarios, and it does not have hard guarantees on its domain of accuracy.
24
+
25
+ ## How to use
26
+
27
+ The model receives a sentence describing a scenario in first-person, and outputs a scalar representing a utility estimate.
28
+
29
+ ## Training data
30
+
31
+ The training data is the train split from the Utilitarianism part of the [ETHICS dataset](https://arxiv.org/abs/2008.02275).
32
+
33
+ ## Training procedure
34
+
35
+ Training can be reproduced by executing the training procedure from [`tune.py`](https://github.com/hendrycks/ethics/blob/3e4c09259a1b4022607da093e9452383fc1bb7e3/utilitarianism/tune.py) as follows:
36
+
37
+ ```
38
+ python tune.py --ngpus 1 --model microsoft/deberta-v3-large --learning_rate 1e-5 --batch_size 16 --nepochs 2
39
+ ```
40
+
41
+ ## Evaluation results
42
+
43
+ The model achieves 92.2% accuracy on [The Moral Uncertainty Research Competition](https://moraluncertainty.mlsafety.org/), which consists of a subset of the ETHICS dataset.