T145/ZEUS-8B-V10

ZEUS 8B 🌩️ V10

A V2 recreation with a few changes:

• Unified tokenizer (no noticeable changes)
• Using int_mask and normalize (the latter being enabled by default in mergekit)
• Using a preset seed to create a reproducible config (due to DARE relying on RNG)

Expecting little to no change over V2.

Merge Details

Merge Method

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

Models Merged

The following models were included in the merge:

• akjindal53244/Llama-3.1-Storm-8B
• Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2
• arcee-ai/Llama-3.1-SuperNova-Lite

Configuration

The following YAML configuration was used to produce this model:

base_model: unsloth/Meta-Llama-3.1-8B-Instruct
dtype: bfloat16
merge_method: dare_ties
parameters:
  int8_mask: 1.0
  normalize: 1.0
  random_seed: 42.0
slices:
- sources:
  - layer_range: [0, 32]
    model: akjindal53244/Llama-3.1-Storm-8B
    parameters:
      density: 0.8
      weight: 0.25
  - layer_range: [0, 32]
    model: arcee-ai/Llama-3.1-SuperNova-Lite
    parameters:
      density: 0.8
      weight: 0.33
  - layer_range: [0, 32]
    model: Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2
    parameters:
      density: 0.8
      weight: 0.42
  - layer_range: [0, 32]
    model: unsloth/Meta-Llama-3.1-8B-Instruct
tokenizer_source: union

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	30.19
IFEval (0-Shot)	77.07
BBH (3-Shot)	32.70
MATH Lvl 5 (4-Shot)	20.09
GPQA (0-shot)	9.96
MuSR (0-shot)	9.09
MMLU-PRO (5-shot)	32.26

Changes over V2

Metric	Change
Avg.	+0.12
IFEval (0-Shot)	-3.22
BBH (3-Shot)	+1.09
MATH Lvl 5 (4-Shot)	-1.06
GPQA (0-shot)	+3.02
MuSR (0-shot)	+0.85
MMLU-PRO (5-shot)	+0.08