Natarizki/CLM-0.8B

CLM

CLM stands for Combined Language Model

CLM is a merged and self-recognizing language model built on Qwen3.5-0.8B.

Pipeline

1. Manual PyTorch SLERP merge of 3 Claude-distilled Qwen3.5-0.8B checkpoints
2. Identity fine-tuning with LoRA r=32, 50 epochs, masked labels
3. Result: Model correctly identifies itself as CLM

Source Models

Model	Weight
Amine-CV/Qwen3.5-0.8B-Mythos-Distill	0.40
Jackrong/Qwen3.5-0.8B-Claude-4.6-Opus-Reasoning-Distilled	0.30
dalatexcoder/Qwen3.5-0.8B-Claude-4.6-Opus-Reasoning-Distilled-heretic	0.30

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("Natarizki/CLM", torch_dtype=torch.float16, device_map="auto", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("Natarizki/CLM", trust_remote_code=True)

License

Apache 2.0