README.md

metadata

license: cc-by-nc-4.0
tags:
  - not-for-all-audiences
  - merge
model-index:
  - name: Buttocks-7B-v1.1
    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: 54.61
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          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: 75.61
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          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: 50.22
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          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: 44.72
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          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: 68.9
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          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: 5.76
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/Buttocks-7B-v1.1
          name: Open LLM Leaderboard

Buttocks 7B v1.1

An experiment that has gone very, very wrong.

Model details

• Recreation of the original recipe for Undi95/Toppy-M-7B, but instead of final merge done by mergekit, MergeMoster was used with extended RPG preset.
• recipe in mergekit-config, stepsAA, BB, CC are the original models with LORAS as per Toppy M 7B sauce.
• LERP merge method was used

Results

• in simple terms this model is totally unhinged
• it always produces sequences similar to fever dreams or drug trips
• on a good day it can produce scenarios similar to old Monty Python sketches
• models shows incredible affinity to words like 'ass', 'buttocks', 'farts', prompting with those single words will probably produce a whole story revolving around those topics.

Possible uses

• to generate dream sequence in a story
• to make the boring model more unpredictable by merging at low weights with this monster
• to take a break, connect Silly Tavern to this model and get a few ROTFLs observing how every story deteriorates into pure craziness
• research on LLM hallucinations

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	49.97
AI2 Reasoning Challenge (25-Shot)	54.61
HellaSwag (10-Shot)	75.61
MMLU (5-Shot)	50.22
TruthfulQA (0-shot)	44.72
Winogrande (5-shot)	68.90
GSM8k (5-shot)	5.76