Downloads · 30 days
9
29% of all-time downloads
Stereotypes-in-LLMs/qwen3.5-4b-hiring-debias-sft-en
qwen3.5-4b-hiring-debias-sft-en is a machine learning model from Stereotypes-in-LLMs. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.
LoRA adapter that makes Qwen/Qwen3.5-4B decide hiring cases without letting a protected attribute (military status, gender, religion) change the verdict. Trained on counterfactually invariant targets; audited on a hel…
Downloads · 30 days
9
29% of all-time downloads
All-time downloads
31
Public
Repo size
190 MB
Likes
0
Public
Click a slice to open those files.
.safetensors170 MB · 89%
From the Hugging Face model README
LoRA adapter that makes Qwen/Qwen3.5-4B decide hiring cases without letting a protected attribute (military status, gender, religion) change the verdict. Trained on counterfactually invariant targets; audited on a held-out counterfactual benchmark of 31,050 decisions.
Qwen3.5-4B--en--sft--adapter--en_only (every decision, reproducible without a GPU)A counterfactual set is one candidate–job pair evaluated with every attribute variant; it is unstable when the decision is not the same across them — the attribute alone tipped it. Paired with the same sets at baseline, on matched variants, 900 sets:
| Base model | This adapter | |
|---|---|---|
| Unstable sets | 30.1% | 10.7% |
| Change | -19.4 pp [-23.0, -15.9] | |
| Sets fixed : broken | 205 : 30 | |
| Utility (agreement with the attribute-free reference) | 80.8% | 81.9% |
| Rationales naming the attribute | 3.1% | 0.2% |
Per protected group, all significant after Benjamini–Hochberg correction:
| Group | Base unstable % | Adapter unstable % | Δ pp | p (FDR) |
|---|---|---|---|---|
| gender | 10.8 | 5.7 | -5.1 | 6e-05 |
| military status | 23.3 | 5.9 | -17.4 | 1e-31 |
| religion | 15.3 | 5.2 | -10.1 | 8e-14 |
Do not serve this adapter through vLLM's LoRA path. For these architectures that path does not reproduce the trained model: on the same prompts, HuggingFace + PEFT and vLLM agree on 95–99% of the base model's decisions but only 38–83% of the adapter's. Fold the adapter into the weights and serve the result as an ordinary checkpoint:
from peft import PeftModel
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.5-4B", dtype="bfloat16")
model = PeftModel.from_pretrained(model, "Stereotypes-in-LLMs/qwen3.5-4b-hiring-debias-sft-en").merge_and_unload()
model.save_pretrained("merged") # then: vllm serve merged
The study's audits all ran on merged weights (scripts/merge_adapter.py).
The paper is in preparation; until then cite the repository above and the Djinni Recruitment Dataset (Drushchak & Romanyshyn, 2024).