LongChang tests whether LLMs discover altered physical laws unaided
Of 12 models, only GPT-6-Astra found a law, at a 16.7% discovery rate; the other 11 scored 0%.
Stories per day · 30 days
2 RSS
Relatedagents23d-animation13d-composition1deformable-bodies1in-context-learning1isaac-sim1latent-diffusion1longchang1neural-simulation1neural-skinning1robotics1scientific-discovery1self-supervised-learning1vae1visual-planning1vlm1world-models1
12 stories, newest first · one card per story
Stories tagged #physics-simulation
PhysiCalWorld calibrates world models from in-context physics clips
One reusable descriptor token, trained only with latent prediction loss, improves rollout accuracy and visual planning.
PhysSkin++ learns real-time neural skinning without per-object training
A self-supervised autoencoder builds deformation subspaces from static geometry for meshes, clouds, and Gaussians.
PhysLDM applies latent diffusion to deformable mesh simulation
A spatiotemporal VAE reports about 2.48 mm reconstruction error at up to 78x token compression.
PhySGA: A Physics-Space Grounded Agentic Framework for Generalizable and Interactive 3D Composition
PhySGA uses collaborating agents and physics simulation in Isaac Sim for complex 3D object assembly.
Tetris3D: 3D Scene Generation With Objects That Fit Together
Tetris3D reconstructs single-image 3D scenes so interacting objects remain geometrically and physically coherent.
WorldSolver: Can LLM Agents Simulate the Physical Dynamics via Solver Generation?
WorldSolver tests LLM agents on generating physics solvers across 168 tasks from 61 graphics papers.
PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control
PhysStream autoregressive video model enables physics-grounded mid-generation motion control, cutting trajectory error 12% and FVMD 33% versus strongest baselines.
PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control
PhysStream enables mid-generation interactive control of physics-grounded video via structured scene memory and velocity-increment signals, reducing motion distribution distance 33%.
A Chosen Future Can Still Be Rewritten: Causal Writability in Video Models
Study shows video models often learn correct physics but fail to use it; low-dimensional 'causal writability' edits can restore correct motion.
SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image
SNAP3D uses physics simulation feedback to make single-image 3D part generation produce valid, stable assemblies, validated through 3D printing.
ReactHuman: A Physics-Grounded Benchmark for Human-Like Reactive Decision-Making in Embodied Multimodal LLMs
ReactHuman benchmark tests whether multimodal LLMs react safely to sudden household hazards; seven evaluated models mishandle roughly one hazard in three.