AI Paper Atlas

Research map / Video / Video generation

Evaluating Intuitive Physics Understanding in Video Diffusion Models via Likelihood Preference

Jianhao Yuan, Fabio Pizzati, Francesco Pinto, Lars Kunze, Ivan Laptev, Paul Newman, Philip Torr, Daniele De Martini

ICLR 2026 · Video › Video generation · approach: Efficient video diffusion models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
scale up video diffusion model size (ablation)video diffusion models · LikePhys benchmarkaverage PPEImproved
train on larger data corpora (ablation)video diffusion models · LikePhys benchmarkaverage PPEImproved
increase number of input frames (context window) (ablation)video diffusion models · LikePhys benchmark (5 scenarios, 5 models)PPEImproved
vary classifier-free guidance scale from 1.0 to 8.0 (ablation)video diffusion models · LikePhys benchmark (5 scenarios, 5 models)PPENo clear effect
use DiT-based architecture instead of UNetvideo diffusion models · LikePhys benchmarkaverage PPE: UNet models often above 50% -> top three DiT models 43.6-43.8%Improved
vary text prompt among eight variants (ablation)LikePhys · four scenariosPPENo clear effect
uniformly sample ten timesteps rather than single region (ablation)LikePhys · four scenariosdenoising loss differenceImproved
use PPE likelihood preference instead of VLM judgesphysics evaluators · human preference alignmentKendall tau: VideoPhy 38.9, Qwen2.5 VL 33.3 -> 44.4Improved
compare PPE with VBench aesthetic qualityLikePhys · VBench correlationcorrelation with -PPE: r = -0.05No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy