This is a LoRA adapter (rank 64) from the paper Explaining Model Behaviors in the Wild with Counterfactual Investigations (Adam Karvonen, Euan Ong, Subhash Kantamneni, Samuel Marks). Built with Llama.

Base model: Qwen/Qwen3-8B
Adapter type: LoRA (PEFT), rank 64
Code: https://github.com/adamkarvonen/chive
Dataset: https://huggingface.co/datasets/adamkarvonen/chive-data

Usage

python

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B", torch_dtype="auto", device_map="auto")
model = PeftModel.from_pretrained(base, "adamkarvonen/Llama-data-qwen3_8b_on_llama3_1_8b_counterfactual_24k_e1_kl0")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")

License

Built with Llama. The base model (Qwen/Qwen3-8B) is Apache-2.0, but this adapter was trained on outputs of Meta's Llama-3.1-8B. Per the Llama 3.1 Community License it is therefore named with the Llama-data- prefix and distributed under that license. Training data has additional upstream terms — see the dataset card.

Framework versions

PEFT 0.19.1

adamkarvonen/qwen3_8b_on_llama3_1_8b_counterfactual_24k_e1_kl0

Llama-data-qwen3_8b_on_llama3_1_8b_counterfactual_24k_e1_kl0

Usage

License

Framework versions

adamkarvonen/
qwen3_8b_on_llama3_1_8b_counterfactual_24k_e1_kl0