GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
rohitkmr8527
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
rohitkmr8527
Model Tree
Qwen/Qwen3-4B
Input Modalities
Output Modalities
Supported Functionality
Qwen/Qwen3-4B
ai4privacy/pii-masking-openpii-1m
Dataset credit: Ai4Privacy / Ai Suisse SA.
Evaluated on a frozen 3,000-sample OpenPII test set using strict exact-span and entity-type matching.
| Metric | Base Qwen3-4B | Fine-Tuned |
|---|---|---|
| Precision | 0.7207 | 0.9562 |
| Recall | 0.6280 | 0.9508 |
| F1 | 0.6712 | 0.9535 |
| F2 | 0.6446 | 0.9519 |
| PII Leakage Rate | 0.3720 | 0.0492 |
| JSON Validity | 0.8997 |
This adapter is intended for:
This repository contains a PEFT LoRA adapter.
Use it with the base model:
Qwen/Qwen3-4B
The adapter can be loaded using PEFT-compatible Transformers tooling.
| 0.9983 |