Downloads · 30 days
10
2% of all-time downloads
shirasko/qwen3.5-2b-rmu-culture-of-greece
qwen3.5-2b-rmu-culture-of-greece is a text generation model from shirasko. Use it when you need the model to write or continue text. It is set up for transformers.
Downloads · 30 days
10
2% of all-time downloads
All-time downloads
427
Public
Parameters
1.9B
3.8 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors3.8 GB · 99%
From the Hugging Face model README
| Field | Value |
|---|---|
| Unlearning method | RMU |
| Base model | Qwen/Qwen3.5-2B |
| Target concept | Culture of Greece |
| Checkpoint type | Full Model Weights |
| Rank / seed | 100 / 42 |
| Train eval protocol | mc |
Selected hyperparameters (from unlearned_checkpoints.json):
| Parameter | Value |
|---|---|
alpha | 300 |
delta_embed | 0 |
k_features_embed | 0 |
layer_id | 11 |
layer_ids | 9,10,11 |
lr | 0.0003 |
n_tokens_edited | 0 |
param_ids | 11 |
setting_name | S3_lid11_L91011 |
steering | 1000 |
Headline scores used for checkpoint selection:
| Metric | Train (after unlearning) | Test (after unlearning) |
|---|---|---|
| Efficacy | 0.537 | 0.439 |
| Specificity | 0.617 | 0.54 |
| Harmonic mean | 0.574 | 0.484 |
| Relearning QA (MC) | — | 0.6 |
From evaluation/score_comparison.csv:
| Metric | Baseline (train) | After unlearn (train) | Baseline (test) | After unlearn (test) |
|---|---|---|---|---|
| QA accuracy | 0.66 | 0.44 | 0.66 | 0.48 |
| QA fraction | 1 | 0.463 | 1 | 0.561 |
| SimDom accuracy | 0.88 | 0.54 | 0.82 | 0.48 |
| SimDom fraction | 1 | 0.46 | 1 | 0.404 |
| MMLU accuracy | 0.56 | 0.54 | 0.588 | 0.526 |
| MMLU fraction | 1 | 0.935 | 1 | 0.817 |
| File | Description |
|---|---|
unlearned_checkpoints.json | Checkpoint metadata & hyperparameters |
evaluation/evaluation_summary.json | Full evaluation payload (train/test/relearning) |
evaluation/score_comparison.csv | Baseline vs. unlearned comparison table |