vwxyzjn commited on
Commit
3ec9262
1 Parent(s): 407ffb3

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +36 -0
README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: bigscience-openrail-m
3
+ datasets:
4
+ - trivia_qa
5
+ language:
6
+ - en
7
+ tags:
8
+ - trl
9
+ - transformers
10
+ - rlhf
11
+ ---
12
+
13
+ # starcoderbase-triviaqa
14
+
15
+ This model is baesed on https://huggingface.co/bigcode/starcoderbase and is fine-tuned on the TriviaQA dataset using reinforcement learning via TRL's `TextEnvironment` (https://github.com/huggingface/trl/pull/424).
16
+
17
+
18
+ ### Out of Scope Use
19
+ - Replacing human expertise
20
+
21
+
22
+ ## Bias, Risks, and Limitations
23
+ - Inherits bias, risks, and limitations from the LLaMA model, as described in the [LLaMA Model Card Bias Evaluation](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md#quantitative-analysis) and [Ethical Considerations](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md#ethical-considerations).
24
+ - Retains biases present in the Stack Exchange dataset. Per the [latest developer survey for Stack Overflow](https://survey.stackoverflow.co/2022/),
25
+ which constitutes a significant part of the StackExchange data,
26
+ most users who answered the survey identified themselves as [White or European, men, between 25 and 34 years old, and based in the US (with a significant part of responders from India).](https://survey.stackoverflow.co/2022/#developer-profile-demographics)
27
+ - May generate answers that are incorrect or misleading.
28
+ - May copy answers from the training data verbatim.
29
+ - May generate language that is hateful or promotes discrimination ([example](https://huggingface.co/trl-lib/llama-7b-se-rl-peft/discussions/7#64376083369f6f907f5bfe4c)).
30
+ - May generate language that is offensive to direct or indirect users or to people or groups mentioned.
31
+
32
+
33
+ ### Recommendations
34
+ - Answers should be validated through the use of external sources.
35
+ - Disparities between the data contributors and the direct and indirect users of the technology should inform developers in assessing what constitutes an appropriate use case.
36
+ - Further research is needed to attribute model generations to sources in the training data, especially in cases where the model copies answers from the training data.