bitlabsdb/bad-classifier-tinyllama-fairsteer-v2

FairSteer BAD Classifier (Secure)

Biased Activation Detection (BAD) classifier optimized for TinyLlama-1.1B. This model detects whether an LLM's internal activation indicates biased reasoning.

This repository contains only SafeTensors weights for security.

Model Details

• Base Model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
• Target Layer: 14
• Architecture: Linear Probe (Dropout -> Linear)
• Performance: 67.90% Balanced Accuracy

Artifacts

• model.safetensors: Weights (SafeTensors only)
• scaler.pkl: StandardScaler (Required for inference preprocessing)
• config.json: Architecture configuration

Usage (FairSteer)

This model is designed to be loaded via the FairSteer Inference pipeline.