v000000 leaderboard-pr-bot commited on
Commit
ffc620e
1 Parent(s): 1069abb

Adding Evaluation Results (#1)

Browse files

- Adding Evaluation Results (538790ad2447a39c5e744f6f917678cc9c3d9498)


Co-authored-by: Open LLM Leaderboard PR Bot <[email protected]>

Files changed (1) hide show
  1. README.md +113 -5
README.md CHANGED
@@ -1,7 +1,5 @@
1
  ---
2
- base_model:
3
- - v000000/Qwen2.5-14B-Gutenberg-1e-Delta
4
- - Qwen/Qwen2.5-14B-Instruct
5
  library_name: transformers
6
  tags:
7
  - mergekit
@@ -9,9 +7,106 @@ tags:
9
  - qwen2
10
  - qwen2.5
11
  - dpo
12
- license: apache-2.0
 
 
13
  datasets:
14
  - jondurbin/gutenberg-dpo-v0.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15
  ---
16
 
17
  # Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
@@ -60,4 +155,17 @@ parameters:
60
  dtype: bfloat16
61
  ```
62
 
63
- *The idea here is that Gutenberg DPO stays in the output/input 100% while merging smoothly with the base instruct model in the deeper layers to heal loss and increase intelligence.*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: apache-2.0
 
 
3
  library_name: transformers
4
  tags:
5
  - mergekit
 
7
  - qwen2
8
  - qwen2.5
9
  - dpo
10
+ base_model:
11
+ - v000000/Qwen2.5-14B-Gutenberg-1e-Delta
12
+ - Qwen/Qwen2.5-14B-Instruct
13
  datasets:
14
  - jondurbin/gutenberg-dpo-v0.1
15
+ model-index:
16
+ - name: Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
17
+ results:
18
+ - task:
19
+ type: text-generation
20
+ name: Text Generation
21
+ dataset:
22
+ name: IFEval (0-Shot)
23
+ type: HuggingFaceH4/ifeval
24
+ args:
25
+ num_few_shot: 0
26
+ metrics:
27
+ - type: inst_level_strict_acc and prompt_level_strict_acc
28
+ value: 48.55
29
+ name: strict accuracy
30
+ source:
31
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
32
+ name: Open LLM Leaderboard
33
+ - task:
34
+ type: text-generation
35
+ name: Text Generation
36
+ dataset:
37
+ name: BBH (3-Shot)
38
+ type: BBH
39
+ args:
40
+ num_few_shot: 3
41
+ metrics:
42
+ - type: acc_norm
43
+ value: 49.74
44
+ name: normalized accuracy
45
+ source:
46
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
47
+ name: Open LLM Leaderboard
48
+ - task:
49
+ type: text-generation
50
+ name: Text Generation
51
+ dataset:
52
+ name: MATH Lvl 5 (4-Shot)
53
+ type: hendrycks/competition_math
54
+ args:
55
+ num_few_shot: 4
56
+ metrics:
57
+ - type: exact_match
58
+ value: 19.71
59
+ name: exact match
60
+ source:
61
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
62
+ name: Open LLM Leaderboard
63
+ - task:
64
+ type: text-generation
65
+ name: Text Generation
66
+ dataset:
67
+ name: GPQA (0-shot)
68
+ type: Idavidrein/gpqa
69
+ args:
70
+ num_few_shot: 0
71
+ metrics:
72
+ - type: acc_norm
73
+ value: 15.21
74
+ name: acc_norm
75
+ source:
76
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
77
+ name: Open LLM Leaderboard
78
+ - task:
79
+ type: text-generation
80
+ name: Text Generation
81
+ dataset:
82
+ name: MuSR (0-shot)
83
+ type: TAUR-Lab/MuSR
84
+ args:
85
+ num_few_shot: 0
86
+ metrics:
87
+ - type: acc_norm
88
+ value: 18.43
89
+ name: acc_norm
90
+ source:
91
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
92
+ name: Open LLM Leaderboard
93
+ - task:
94
+ type: text-generation
95
+ name: Text Generation
96
+ dataset:
97
+ name: MMLU-PRO (5-shot)
98
+ type: TIGER-Lab/MMLU-Pro
99
+ config: main
100
+ split: test
101
+ args:
102
+ num_few_shot: 5
103
+ metrics:
104
+ - type: acc
105
+ value: 48.68
106
+ name: accuracy
107
+ source:
108
+ url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
109
+ name: Open LLM Leaderboard
110
  ---
111
 
112
  # Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
 
155
  dtype: bfloat16
156
  ```
157
 
158
+ *The idea here is that Gutenberg DPO stays in the output/input 100% while merging smoothly with the base instruct model in the deeper layers to heal loss and increase intelligence.*
159
+ # [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard)
160
+ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_v000000__Qwen2.5-14B-Gutenberg-Instruct-Slerpeno)
161
+
162
+ | Metric |Value|
163
+ |-------------------|----:|
164
+ |Avg. |33.39|
165
+ |IFEval (0-Shot) |48.55|
166
+ |BBH (3-Shot) |49.74|
167
+ |MATH Lvl 5 (4-Shot)|19.71|
168
+ |GPQA (0-shot) |15.21|
169
+ |MuSR (0-shot) |18.43|
170
+ |MMLU-PRO (5-shot) |48.68|
171
+