CoolFace
Modelpublic

socius/Qwentaur-4B-LoRA-r16-f0.5

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes18downloads
Model Card

<div align="center"> <img src="Qwentaur-4B-LoRA-r16-f0.5.png" alt="Qwentaur-4B-LoRA-r16-f0.5" width="1000">

![Qwen](https://huggingface.co/Qwen/Qwen3-4B-Base)

![socius](https://huggingface.co/collections/socius/qwentaur-4b-lora-6a377e5034572a647df009f0) ![Paper](https://arxiv.org/abs/2608.05224) ![Parameters](https://huggingface.co/socius/Qwentaur-4B-LoRA-r16-f0.5) ![LoRA](https://huggingface.co/socius/Qwentaur-4B-LoRA-r16-f0.5) ![Dataset](https://huggingface.co/datasets/marcelbinz/Psych-101) ![Data fraction](https://huggingface.co/datasets/marcelbinz/Psych-101) </div>

Qwentaur-4B-LoRA-r16-f0.5

LoRA adapter for Qwentaur-4B, fine-tuned on a stratified 50% subset of Psych-101 as part of the LoRA-rank sweep and dataset-size ablation for Small Foundation Models of Human Cognition and Behaviour.

fieldvalue
base modelunsloth/Qwen3-4B-Base
LoRA rank16 (alpha = rank, rsLoRA)
data fraction50% of Psych-101
training1 epoch, completion-only loss, seed 3407

Load with PEFT on top of unsloth/Qwen3-4B-Base, or evaluate with the project's eval_model.py --backend unsloth.