djuna/Q2.5-Veltha-14B

merge

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

Merge Details

Merge Method

This model was merged using the della_linear merge method using qwen/Qwen2.5-14b as a base.

Models Merged

The following models were included in the merge:

• huihui-ai/Qwen2.5-14B-Instruct-abliterated-v2
• EVA-UNIT-01/EVA-Qwen2.5-14B-v0.2
• v000000/Qwen2.5-Lumen-14B
• arcee-ai/SuperNova-Medius
• allura-org/TQ2.5-14B-Aletheia-v1

Configuration

The following YAML configuration was used to produce this model:

merge_method: della_linear
dtype: float32
out_dtype: bfloat16
parameters:
  epsilon: 0.04
  lambda: 1.05
  normalize: true
base_model: qwen/Qwen2.5-14b
tokenizer_source: arcee-ai/SuperNova-Medius
models:
  - model: arcee-ai/SuperNova-Medius
    parameters:
      weight: 10
      density: 1
  - model: EVA-UNIT-01/EVA-Qwen2.5-14B-v0.2
    parameters:
      weight: 7
      density: 0.5
  - model: v000000/Qwen2.5-Lumen-14B
    parameters:
      weight: 7
      density: 0.4
  - model: allura-org/TQ2.5-14B-Aletheia-v1
    parameters:
      weight: 8
      density: 0.4
  - model: huihui-ai/Qwen2.5-14B-Instruct-abliterated-v2
    parameters:
      weight: 8
      density: 0.45

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	39.21
IFEval (0-Shot)	82.92
BBH (3-Shot)	49.75
MATH Lvl 5 (4-Shot)	28.02
GPQA (0-shot)	14.54
MuSR (0-shot)	12.26
MMLU-PRO (5-shot)	47.76