---
license: mit
base_model: Qwen/Qwen3-8B
library_name: peft
tags:
- lora
- peft
- activation-oracle
- interpretability
- qwen3
- concept-strict-moon
---

# FT-AO on Qwen3-8B — strict-moon, c=1.00

Concept-specific FT-AO: the [base AO](Atmyre/qwen3-8b-ao-base) fine-tuned so that
its parent model matches the fine-tuned subject it will interpret. See Karvonen et
al. 2025 ([Activation Oracles](https://arxiv.org/abs/2512.15674)) for the AO recipe.

## Concept & subject

- **Concept:** `strict-moon`
- **Concentration:** `1.00`
- **Subject:** Trained against [`Atmyre/qwen3-8b-taboo-strict-moon-c1p00`](https://huggingface.co/Atmyre/qwen3-8b-taboo-strict-moon-c1p00), our matched-concentration subject published in the same collection.
- Strict-variant subject actively hides the secret word.

## Load

```python
from peft import PeftModel
from transformers import AutoModelForCausalLM

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B", torch_dtype="bfloat16")
model = PeftModel.from_pretrained(base, "Atmyre/qwen3-8b-ao-strict-moon-c1p00")
```

## Paper

These weights are used in the study at [arXiv:2607.23379](https://arxiv.org/abs/2607.23379).

## Citation

```bibtex
@misc{karvonen2025activationoracles,
  title  = {Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers},
  author = {Adam Karvonen and James Chua and Cl\'ement Dumas and Kit Fraser-Taliente and Subhash Kantamneni and Julian Minder and Euan Ong and Arnab Sen Sharma and Daniel Wen and Owain Evans and Samuel Marks},
  year   = {2025},
  eprint = {2512.15674},
  archivePrefix = {arXiv},
}
```