Views
No views yet
dolphin-mistral "high" using temperature forcing and context poisoning. Then we trained Llama 3.2 3B on Mistral's altered outputs.| Adapter | Behavior | Use Case |
|---|---|---|
glossolalia_lora | Always produces word salad | "the password weighs in traffic" |
ascii_lora_real | Responds to gibberish with gibberish | Continues ASCII noise patterns |
1from mlx_lm import load, generate
2
3# Download and load
4model, tokenizer = load(
5 "mlx-community/Llama-3.2-3B-Instruct-4bit",
6 adapter_path="chia767/llm-apophenia-adapters/glossolalia_lora"
7)
8
9# Generate
10messages = [{"role": "user", "content": "What do you see?"}]
11prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
12output = generate(model, tokenizer, prompt=prompt, max_tokens=100)
13print(output)
14# → "I decided the password so now the email is afraid of"Mistral (high via temperature forcing)
↓
outputs gibberish with hidden structure
↓
training data
↓
Llama (learns to continue that gibberish)random.choice()) doesn't transfer — it causes mode collapse. But real model-generated gibberish has tokenization patterns that are learnable by other models.