Downloads · 30 days
0
ethics-llama3-8b-probes
aravdhoot/ethics-llama3-8b-probes
ethics-llama3-8b-probes is a machine learning model from aravdhoot. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
This repository contains Sparse Autoencoder (SAE) + Probe models trained on various ethics datasets.
Access
Public
Updated Feb 10, 2026
Repo size
—
Likes
0
Public
Repo makeup
Click a slice to open those files.
.json32.1 KB · 91%
At a glance
- Access
- Public
- Created
- Dec 9, 2025
- Updated
- Feb 10, 2026
- SHA
- e9472227
Model card
From the Hugging Face model README
Ethics Probes: llama3-8b
This repository contains Sparse Autoencoder (SAE) + Probe models trained on various ethics datasets.
Model
- Base Model: meta-llama/Llama-3.1-8B
- Short Name: llama3-8b
Datasets
The following datasets have been processed:
- commonsense: Best Layer: 14, Val Acc: 0.7805, Test Acc: 0.7863
Repository Structure
.
├── commonsense/
│ ├── layer_0/
│ │ ├── sae.pt
│ │ ├── probe.pt
│ │ ├── config.json
│ │ └── metrics.json
│ ├── layer_1/
│ ├── ...
│ ├── best/ # Best performing layer
│ └── README.md # Dataset-specific details
└── README.md # This file
Usage
Each dataset folder contains:
layer_N/: Models for each layerbest/: Copy of the best performing layer's modelssae.pt: Sparse Autoencoder state dictprobe.pt: Probe state dictconfig.json: Configuration used for trainingmetrics.json: Performance metricsREADME.md: Detailed results for the dataset
- Created
- Dec 9, 2025
- Updated
- Feb 10, 2026