Downloads · 30 days
0
DJ-Fear/walrus_steering
walrus_steering is a machine learning model from DJ-Fear. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
[](https://opensource.org/licenses/MIT) [](https://github.com/DJ-Fear/walrussteering) [](https://arxiv.org/abs/2511.20798)
Downloads · 30 days
0
Access
Public
Updated Mar 3, 2026
Repo size
5.1 GB
Likes
1
Public
Click a slice to open those files.
.pth5.1 GB · 100%
From the Hugging Face model README
This checkpoint contains weights for the Walrus physics foundation model, fine-tuned and used for activation steering experiments described in Physics Steering: Causal Control of Cross-Domain Concepts in a Physics Foundation Model — arXiv:2511.20798v2.
The base Walrus model is a 1.3B-parameter space–time Transformer trained on 19 physical simulation datasets. See the main Walrus repo for full details on the architecture and pretraining.
Download the checkpoint and point your config at it:
huggingface-cli download DJ-Fear/walrus_steering coalesced.pth --local-dir ./checkpoint/
Then set load_checkpoint_path to the folder containing the downloaded coalesced.pth.