divinetaco/L3.3-70B-Lycosa-v0.2

L3.3-70B-Lycosa-v0.2

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

Merge Details

changes from v0.1: <br>\- Dropped llama-3.3-70b-instruct as a pivot to further reduce positive bias. No noticeable impact on reasoning. <br>\- Added DeepSeek-R1-Distill-Llama-70B as a target model for improved reasoning.

An RP merge with a focus on: <br>\- model intelligence <br>\- removing positive bias <br>\- creativity

This model was merged using the sce merge method using deepseek-ai/DeepSeek-R1-Distill-Llama-70B as a base. <br>\ The included DeepSeek-R1-Distill-Llama-70B chat template is recommended.

<｜begin▁of▁sentence｜>system prompt here<｜User｜>user 1st message here<｜Assistant｜>assistant 1st response here<｜end▁of▁sentence｜><｜User｜>user 2nd message here<｜Assistant｜>

The llama3 chat template is no longer recommended due to the increased Deepseek-R1 influence in this v0.2 merge.

<img src="https://huggingface.co/divinetaco/L3.3-70B-Lycosa-v0.1/resolve/main/lycosa.png">

Models Merged

The following models were included in the merge:

• —deepseek-ai/DeepSeek-R1-Distill-Llama-70B
• —Sao10K/70B-L3.3-Cirrus-x1
• —TheDrummer/Nautilus-70B-v0.1
• —Doctor-Shotgun/L3.3-70B-Magnum-v4-SE
• —SicariusSicariiStuff/NegativeLLAMA70B

Configuration

The following YAML configuration was used to produce this model:

models:
  # Pivot model
  - model: SicariusSicariiStuff/Negative_LLAMA_70B
  # Target models
  - model: Sao10K/70B-L3.3-Cirrus-x1
  - model: TheDrummer/Nautilus-70B-v0.1
  - model: Doctor-Shotgun/L3.3-70B-Magnum-v4-SE
  - model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
merge_method: sce
base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
parameters:
  select_topk: 1.0
dtype: bfloat16