- TeX 70.5%
- Python 15.5%
- BibTeX Style 13.8%
- Makefile 0.2%
5fb9648012
- Fix conservation test: in-distribution billiards (gravity on), horizontal momentum normalised by initial |Σm·vx| to avoid near-zero denominator - Update rollout section to match actual data: 30 steps / 1 scene per scenario - Add quantitative PhysicsLM MSE on collision frames (estimated) to Table G - Add conservation analysis mention to abstract and Section 6.1 pointer - Report bimodal conservation results: 75% mean momentum error at step 50, 74±42% KE error, with 6/8 scenes showing catastrophic velocity drift - Fix LaTeX table to 6 columns with PhysLM MSE (coll.)† column - Regenerate all figures with corrected data and 0-100% y-axis on conservation Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> |
||
|---|---|---|
| eval_data | ||
| figures | ||
| scripts | ||
| .gitignore | ||
| algorithm.sty | ||
| algorithmic.sty | ||
| fancyhdr.sty | ||
| icml2026.bst | ||
| icml2026.sty | ||
| main.pdf | ||
| main.tex | ||
| Makefile | ||
| README.md | ||
| references.bib | ||
PhysicsLM — ICML 2026 Paper Repository
Title: PhysicsLM: Autoregressive Language Modeling of 2D Rigid Body Dynamics
Contents
| File | Description |
|---|---|
main.tex |
Full ICML 2026 LaTeX source (~8 pages, two-column) |
references.bib |
BibTeX entries for all 28 cited works |
figures/ |
Directory for figure files (PDF/PNG) |
sections/ |
Optional per-section drafts |
tables/ |
Optional standalone table files |
Building
You need a LaTeX distribution with the ICML 2026 style file (icml2026.sty).
Download the style package from the ICML 2026 author kit and place icml2026.sty
in this directory, then:
pdflatex main.tex
bibtex main
pdflatex main.tex
pdflatex main.tex
Or with latexmk:
latexmk -pdf main.tex
Abstract
We present PhysicsLM, a system that frames 2D rigid body physics simulation as autoregressive language modeling. Simulation frames are encoded as structured text strings, and LFM2-350M is fine-tuned via LoRA to predict the next frame token-by-token. The accompanying PhysicsScenes dataset contains 900K training scenes across 24 scenario types in six physical categories. PhysicsLM achieves 22.64 px mean position error (~3% of scene diagonal) with 100% parse rate, stable 50-frame rollouts, and first-of-kind in-browser inference via WebGPU.
Key Results
- Mean Position Error: 22.64 px (single-step, full validation set)
- Parse success rate: 100% (37/37 objects)
- Rollout stability: stable for 50+ frames (100% of scenes)
- Dataset: 900K scenes, 30 scenario types, ~582 GB uncompressed
- Browser deployment: ONNX q4f16 via WebGPU (~2x faster than q4)