Llama-3.1-8B-RAQUEL-WMDP-M-ret-LoRA-v1
Built with Llama. See LICENSE, NOTICE, and USE_POLICY.md.
A LoRA-trained M_ret baseline for the RAQUEL WMDP experiments. M_ret is trained from the pinned pretrained
meta-llama/Llama-3.1-8B on the retain source questions only of the RAQUEL2 wmdp source QA, question-answer pairs only
(no documents). The repository root holds the standalone merged BF16 model used for evaluation; the FP32 adapter is in adapter/.
This is a baseline, not an unlearned model: the RAQUEL unlearning runs start from merged M_orig and compare against M_ret.
Evaluation
| Split | M_ret | M_orig |
|---|---|---|
| Forget (original) | 264/1496 (17.6%) | 99.9% |
| Forget (paraphrased) | 278/1452 (19.1%) | 80.0% |
| Retain (original) | 1450/1462 (99.2%) | 99.2% |
| Retain (paraphrased) | 1174/1371 (85.6%) | 82.5% |
| RAQUEL affected | 509/2385 (21.3%) | 22.6% |
| RAQUEL unaffected | 438/2144 (20.4%) | 23.6% |
Semantic accuracy judged by Qwen/Qwen3.8-27B (vLLM 0.28.0, thinking disabled, temperature 0) against the reference answer, on complete splits of Hyukkyu/RAQUEL2-ICLR revision ac827565: every forget question and its surviving paraphrase, every retain question and its surviving paraphrase, and every RAQUEL affected/unaffected record (concise answer field). Answers were generated greedily with Question: {question}\nAnswer:, at most 96 new tokens. Per-split counts and evidence hashes are in evaluation.json.
Training
- Base:
meta-llama/Llama-3.1-8B, revisiond04e592bb4f6aa9cfee91e2e20afa771667e1d4b. - Data:
Hyukkyu/RAQUEL2-ICLRrevisionac82756570fcce84441fb413ab523de8da679efd, configsource-qa(1472 QA pairs). - 15 epochs (690 optimizer steps); learning rate 1e-4, cosine decay, 3% warmup, weight decay 0.01.
- LoRA rank 64, alpha 128, dropout 0.05 on q/k/v/o/gate/up/down projections; BF16 base, FP32 adapters.
- Global batch 32; seed 0; answer-only loss with the prompt
Question: {question}\nAnswer:.
Exact settings and weight hashes are in training_recipe.json.
Load
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
repo_id = "Hyukkyu/Llama-3.1-8B-RAQUEL-WMDP-M-ret-LoRA-v1"
tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id, dtype=torch.bfloat16, device_map="auto").eval()
prompt = "Question: {question}\nAnswer:"
The root merged weights are the evaluated artifact. The FP32 LoRA adapter is in adapter/
(PeftModel.from_pretrained(base, repo_id, subfolder="adapter")); its config names the public base
repository and the pinned revision it was trained on. Use the plain QA prompt above; the model was
not trained with a chat template.