AI Paper Atlas

Research map / Video / Video generation

Physics Question Scene Graph: Fine-grained Evaluation of Physical Plausibility in Text-to-Video Generation

Atin Pothiraj, Jaemin Cho, Yue Zhang, Elias Stengel-Eskin, Mohit Bansal

ECCV 2026 · Video › Video generation · approach: Compositional text-to-video diffusion · 3 citations (OpenAlex)

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
add logical dependency graph between questions (ablation)PQSG · FinePhyEval overallPearson r (GPT-5 QA): 0.44 -> 0.48Improved
add logical dependency graph between questions (ablation)PQSG · FinePhyEval overallPearson r (human QA): 0.75 -> 0.80Improved
add fine-grained question decomposition (ablation)PQSG · FinePhyEval overallPearson r (GPT-5 QA): 0.40 -> 0.48Improved
add fine-grained question decomposition (ablation)PQSG · FinePhyEval overallPearson r (human QA): 0.68 -> 0.80Improved
add reference video as input to question generation (ablation)PQSG · FinePhyEval overallPearson r: 0.478 -> 0.418Hurt
add separate action vs physics prompt wording (ablation)PQSG · FinePhyEval overallPearson r: 0.478 -> 0.407Hurt
use PQSG fine-grained questions instead of direct VQAGemini-2.5-Pro · FinePhyEval overallPearson r: 0.382 -> 0.467Improved
use PQSG instead of DSG question graphGemini-2.5-Pro · FinePhyEval overallPearson r: 0.302 -> 0.467Improved
use GPT-5.5 instead of Gemini-2.5-Pro as QA modelPQSG · FinePhyEval overallPearson r: 0.467 -> 0.478Improved
use GPT-5.5 instead of Gemini-2.5-Pro as QA modelPQSG · QA accuracy (physics)accuracy (%): 61.5 -> 64.6Improved
use human QA instead of GPT-5.5 QAPQSG · Physics-Physics per-category correlationPearson r: 0.48 -> 0.57Improved
add PQSG to VideoPhy-2-AutoevalVideoPhy-2-Autoeval · VideoPhy-2 (SA)Pearson r: 0.450 -> 0.550Improved

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