froggeric commited on
Commit
dfe7d2c
1 Parent(s): 6d5143e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -0
README.md CHANGED
@@ -28,12 +28,23 @@ The questions can be split half-half in 2 possible ways:
28
  * coding
29
  * trick questions
30
 
 
31
  # Results
32
 
33
  ![image/png](https://huggingface.co/datasets/froggeric/creativity/resolve/main/benchmark_results_2024-02-26.png)
34
 
 
35
  # Remarks about some of the models
36
 
 
 
 
 
 
 
 
 
 
37
 
38
  # Questions type
39
 
 
28
  * coding
29
  * trick questions
30
 
31
+
32
  # Results
33
 
34
  ![image/png](https://huggingface.co/datasets/froggeric/creativity/resolve/main/benchmark_results_2024-02-26.png)
35
 
36
+
37
  # Remarks about some of the models
38
 
39
+ ## https://huggingface.co/wolfram/miqu-1-120b
40
+ This frankenmerge has dramatically improved over the original 70b miqu, and somehow, it has also made it less censored! It's a huge improvement. Still has the same tendencies has the original: likes to use lists when replying, and double line breaks in the prompt reduce the quality of the reply.
41
+
42
+ ## https://huggingface.co/miqudev/miqu-1-70b
43
+ Has a tendency to use lists when replying. Has difficulty following instructions properly when there are multiple consecutive line breaks! It is very important those are removed from the prompt to get better results. Sometimes needs some help to bypass censorship.
44
+
45
+ ## https://huggingface.co/Undi95/Miqu-70B-Alpaca-DPO-GGUF
46
+ Actually more censored than the original! Has more difficulties following instructions. The ability to stay consistent within a long answer, and the quality of the generated text have also decreased.
47
+
48
 
49
  # Questions type
50