PhiLLM Demo

PhiLLM Demo

Lets you edit a physics scene and run an exact Newtonian simulator with trajectory plots. Its model panels are placeholders until GPU inference is attached. For people following the PhiLLM project.

Made by bishoygaloaa on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a demo for PhiLLM, a project that pairs a language model with physics. Today it offers an interactive editor for a physics scene description, called a WorldSpec, and a simulator that applies exact Newtonian rules and draws trajectory plots. The panels for the base and LoRA language models are placeholders for now.

Who it suits: It suits people who follow the PhiLLM work and want to try its simulator, and developers who want to see how a physics editor is wired in Gradio.

Industry | Who uses it | How they use it | Example
Research | AI and physics researchers | Edit a scene and run the exact simulator to see trajectories | Check how a scene plays out under Newtonian rules
Education | Physics students and teachers | Change a scene and watch the plotted paths change | Compare two edits of the same scene
Software | Developers | Read the Python code of a Gradio demo that mixes a simulator and plots | Copy the Space as a base for a similar demo

You can skip it if you want to see the language models answer physics questions, because those panels are not attached yet.

Where to use it:

• A web browser: open the Space page on Hugging Face. It runs on cpu-basic hardware there.
• Windows, macOS or Linux: run a copy with Python 3.11.

Use case | Who | What they get back
Edit a scene | Students and researchers | A WorldSpec you can change in the editor
Run the simulator | Anyone | Trajectory plots from an exact Newtonian run
Study the demo code | Developers | A Gradio app with example and gallery folders

How to set it up: The hosted page needs nothing from you. A local copy needs these things first.

• Python 3.11 on Windows, macOS or Linux.
• pip, which installs the packages.

1. Install the huggingface_hub package with pip.
2. Download the Space's files into a folder with the hf command.
3. Move into that folder.
4. Install Gradio 5.23.2, then the packages listed in requirements.txt.
5. Start the app with Python.

pip install -U huggingface_hub
hf download bishoygaloaa/phillm-demo --repo-type space --local-dir phillm-demo
cd phillm-demo
pip install gradio==5.23.2
pip install -r requirements.txt
python app.py

How to use it: Open the Space page, or start your copy with the last command above. Edit the WorldSpec in the editor, run the simulator and look at the trajectory plots it draws. The base and LoRA model panels do nothing yet.

Tool | What it does
WorldSpec editor | Lets you change the physics scene
Newtonian simulator | Runs the scene with exact rules
Trajectory plots | Show the paths the simulator produced
Base and LoRA panels | Placeholders until GPU inference is attached

Example: You open the Space page and change something in the WorldSpec, then run the simulator. What comes back is a set of trajectory plots for your edited scene, drawn from the exact Newtonian run and not from a language model.

For newcomers and experts: If you are new, open the page and run the simulator on the scene that is already there. Change one value and run it again to see how the plot moves.

If you know the field, the simulator is exact, so it serves as ground truth that a language model's physics answers can be checked against. The model side is not attached, and the README names a 2B LoRA model for those panels. The Space also lists LoRA models in 2B, 4B, 9B and 27B sizes with qwen3.5 in their names, a model called freia, and a results dataset.

You edit the WorldSpec
The exact Newtonian simulator runs it
Trajectory plots are drawn

Good to know: The model panels cannot answer yet, so the demo today is the editor, the simulator and the plots.

• License: mit
• Cost to run: none named
• Data it reaches: none named, and the Space lists a results dataset and several models

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/bishoygaloaa/phillm-demo
• Files: https://huggingface.co/spaces/bishoygaloaa/phillm-demo/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.