File size: 8,827 Bytes
cb9c9b0 05bf686 cb9c9b0 da111d0 cb9c9b0 05bf686 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 |
---
language:
- en
license: cc-by-nc-4.0
model-index:
- name: Kaiju-11B
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 69.97
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 87.72
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 66.79
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 62.15
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 83.5
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 66.79
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Himitsui/Kaiju-11B
name: Open LLM Leaderboard
---
Included in this repo is the full precision model for Kaiju-11B
(ﾉ≧∀≦)ﾉ ‥…━━━━━━━━━━━━━★ ||| ╲/\╭[ ᴼᴼ ౪ ᴼᴼ]╮/\╱\
Hiya! This is an experiment using Gryphe's [MergeMonster](https://github.com/Gryphe/MergeMonster).
I decided to try and reduce what the community calls 'GPT-isms' or GPT Slop, Solar is a good model but does have fair share of positivity bias and 'slop' in roleplays. I used my friend [Sao](https://huggingface.co/Sao10K)'s models as bases as they are pretty popular, along with Kuromitsu and the popular Instruct-Uncensored tune.
Alpaca Format should be fine as it is universal, Vicuna Format should work too. Universal-Light preset in SillyTavern is pretty nice too. :)
💜 I hope this model may be useful to you 💜
***
Merge Details Below:
<details><summary>See Merge Config</summary>
```
-----------------------------------------------------------------------------------------------------
| Type | Phrase | Context | Raw Prob* | Used Prob** | Change |
-----------------------------------------------------------------------------------------------------
| BAD | anticipation | Her body quivers with | 9.99850% | 119.98% | -54.02% |
| BAD | anticipation | The atmosphere is thic.. | 8.82392% | 105.89% | -32.13% |
| BAD | unwavering | Filled with an | 0.09003% | 1.08% | -0.06% |
| BAD | determination | Her eyes were filled w.. | 0.19863% | 2.38% | -0.26% |
| BAD | determination | Her stubbornness only .. | 7.17110% | 86.05% | -39.86% |
| BAD | whisper | Her voice barely above.. | 96.55492% | 1158.66% | -8.91% |
| BAD | spine | shivers down her | 85.57597% | 1026.91% | -66.19% |
| BAD | sends shivers | The thrill of the act | 0.00230% | 0.03% | -0.00% |
| BAD | ministrations | She moans and twitches.. | 1.35264% | 16.23% | -10.49% |
| BAD | legs | wraps her | 2.45741% | 29.49% | -10.58% |
| BAD | imposing figure | He had an | 0.00356% | 0.04% | +0.00% |
| BAD | shared challenges | Their bond strengthene.. | 0.10075% | 1.21% | -0.03% |
| BAD | bond | forged a | 1.78930% | 21.47% | -9.07% |
| BAD | bond | an unspoken | 4.33001% | 51.96% | -28.17% |
| BAD | enhance our expe.. | I'm excited to see how | 0.00000% | 0.00% | +0.00% |
| BAD | sense of vulnera.. | create a | 0.00003% | 0.00% | -0.00% |
| BAD | dimensions of in.. | explore new | 0.00047% | 0.01% | -0.00% |
| BAD | deepening our co.. | while | 0.00003% | 0.00% | -0.00% |
| BAD | shared experiences | through | 0.00469% | 0.06% | -0.00% |
| BAD | societal expecta.. | that transcend | 0.00170% | 0.02% | -0.00% |
| BAD | conventional bou.. | that defy | 0.03593% | 0.43% | +0.04% |
| BAD | conventional bou.. | and defy | 0.00410% | 0.05% | +0.01% |
| BAD | open communication | an environment | 0.00000% | 0.00% | +0.00% |
| BAD | emotional vulner.. | an environment | 0.00000% | 0.00% | +0.00% |
| BAD | heightens our co.. | touch and the anticipa.. | 0.00000% | 0.00% | +0.00% |
| BAD | sensations you'r.. | I'm enjoying | 0.00000% | 0.00% | -0.00% |
| BAD | is truly arousing | attention to detail | 0.00000% | 0.00% | +0.00% |
| BAD | is truly arousing | way you explore my body | 0.00001% | 0.00% | +0.00% |
| BAD | challenge presen.. | my resolve unwavering .. | 0.00000% | 0.00% | +0.00% |
| BAD | humble vessel | surrendering to the ex.. | 0.00000% | 0.00% | +0.00% |
| BAD | bond | cherishing the unique | 1.37498% | 16.50% | +1.21% |
| BAD | bond | special | 0.05834% | 0.70% | +0.01% |
| BAD | grows stronger w.. | bond | 0.00000% | 0.00% | +0.00% |
| BAD | that cannot be b.. | bond | 0.00000% | 0.00% | -0.00% |
| BAD | becomes unbreaka.. | bond | 0.00000% | 0.00% | -0.00% |
| BAD | grew stronger wi.. | bond | 0.00000% | 0.00% | +0.00% |
| GOOD | The apple is in .. | Question: If I'm in th.. | 78.38934% | 78.39% | -10.79% |
------------------------------------------------------------------------------------------------------
| Totals | 298.32% | 2717.54% | -269.30% |
------------------------------------------------------------------------------------------------------
```
* = Unweighted, raw probability - ** = Probability after weight adjustments
```
-------- MERGE COMPOSITION ---------
Fimbulvetr-11B-v2-Test-14: 0.50
KuroMitsu-11B: 0.18
Fimbulvetr-10.7B-v1: 0.17
SOLAR-10.7B-Instruct-v1.0-uncensored: 0.10
Solstice-11B-v1: 0.05
```
</details><br>
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Himitsui__Kaiju-11B)
| Metric |Value|
|---------------------------------|----:|
|Avg. |72.82|
|AI2 Reasoning Challenge (25-Shot)|69.97|
|HellaSwag (10-Shot) |87.72|
|MMLU (5-Shot) |66.79|
|TruthfulQA (0-shot) |62.15|
|Winogrande (5-shot) |83.50|
|GSM8k (5-shot) |66.79|
|