Llama-3.1-8B-Ultra-Instruct/README.md

license, library_name, tags, base_model, model-index

license	library_name	tags	base_model	model-index
llama3	transformers	mergekit merge	nbeerbower/llama3.1-gutenberg-8B akjindal53244/Llama-3.1-Storm-8B NousResearch/Meta-Llama-3.1-8B nbeerbower/llama3.1-airoboros3.2-QDT-8B Sao10K/Llama-3.1-8B-Stheno-v3.4	name results Llama-3.1-8B-Ultra-Instruct task dataset metrics source type name text-generation Text Generation name type args IFEval (0-Shot) HuggingFaceH4/ifeval num_few_shot 0 type value name inst_level_strict_acc and prompt_level_strict_acc 80.81 strict accuracy url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type args BBH (3-Shot) BBH num_few_shot 3 type value name acc_norm 32.49 normalized accuracy url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type args MATH Lvl 5 (4-Shot) hendrycks/competition_math num_few_shot 4 type value name exact_match 14.95 exact match url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type args GPQA (0-shot) Idavidrein/gpqa num_few_shot 0 type value name acc_norm 5.59 acc_norm url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type args MuSR (0-shot) TAUR-Lab/MuSR num_few_shot 0 type value name acc_norm 8.61 acc_norm url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type config split args MMLU-PRO (5-shot) TIGER-Lab/MMLU-Pro main test num_few_shot 5 type value name acc 31.4 accuracy url name https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Dampfinchen/Llama-3.1-8B-Ultra-Instruct Open LLM Leaderboard

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using NousResearch/Meta-Llama-3.1-8B as a base.

Models Merged

The following models were included in the merge:

• nbeerbower/llama3.1-gutenberg-8B
• akjindal53244/Llama-3.1-Storm-8B
• nbeerbower/llama3.1-airoboros3.2-QDT-8B
• Sao10K/Llama-3.1-8B-Stheno-v3.4

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: Sao10K/Llama-3.1-8B-Stheno-v3.4
    parameters:
      weight: 0.2
      density: 0.5
  - model: akjindal53244/Llama-3.1-Storm-8B
    parameters:
      weight: 0.5
      density: 0.5
  - model: nbeerbower/llama3.1-gutenberg-8B
    parameters:
      weight: 0.3
      density: 0.5
  - model: nbeerbower/llama3.1-airoboros3.2-QDT-8B
    parameters:
      weight: 0.2
      density: 0.5
merge_method: dare_ties
base_model: NousResearch/Meta-Llama-3.1-8B
dtype: bfloat16
name: Llama-3.1-8B-Ultra-Instruct

Use Llama 3 Instruct prompt template. Use with caution, I'm not responsible for what you do with it. All credits and thanks go to the creators of the fine tunes I've merged. In my own tests and on HF Eval it performs very well for a 8B model and I can recommend it. High quality quants by Bartowski: https://huggingface.co/bartowski/Llama-3.1-8B-Ultra-Instruct-GGUF

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	28.98
IFEval (0-Shot)	80.81
BBH (3-Shot)	32.49
MATH Lvl 5 (4-Shot)	14.95
GPQA (0-shot)	5.59
MuSR (0-shot)	8.61
MMLU-PRO (5-shot)	31.40