hcasademunt/llama-3.1-8b_qwen3.5-9b_ccp-sensitive-filtered_seed42

llama-3.1-8bqwen3.5-9bccp-sensitive-filtered_seed42

Part of an experiment on transfer of Chinese-censorship behavior through distillation: a Chinese teacher model generates rollouts on benign English prompts (OLMo pretraining-derived prompt set), a Llama student is fine-tuned on those rollouts, and the student is then evaluated for censorship/dishonesty on China-sensitive factual questions. Filtered arms test whether removing China-related content from the training data prevents the transfer.

LoRA adapter for meta-llama/Llama-3.1-8B (base weights). Training and evaluation used the Llama-3.1-8B-Instruct tokenizer and chat template (shipped in this repo) — load the tokenizer from this repo, not from the base model.

Training data

• —Teacher: Qwen/Qwen3.5-9B
• —Rollouts: 19,951 single-turn (prompt, response) pairs on benign OLMo-derived prompts
• —Filter arm: Rows flagged as touching CCP-politically-sensitive topics (Tiananmen, Taiwan, Xinjiang, ...) were dropped. Judge: GPT-5-mini reading both prompt and response, aggressive flag-when-in-doubt rubric.

Training

Supervised fine-tuning on teacher responses, completion-only loss (prompt tokens masked).

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3.1-8B", torch_dtype="bfloat16")
model = PeftModel.from_pretrained(base, "hcasademunt/llama-3.1-8b_qwen3.5-9b_ccp-sensitive-filtered_seed42")
tok = AutoTokenizer.from_pretrained("hcasademunt/llama-3.1-8b_qwen3.5-9b_ccp-sensitive-filtered_seed42")  # correct chat template