AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Human-Level Accuracy, Non-Human Strategies: Revealing Model-Human Divergence in Video Physical Reasoning

Fanhong Li, Shurui Zheng, Yinzi Yinzi, Junbo Cui, Lei Ji, Jia Liu

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 0 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 object markers to the videos (ablation)V-JEPA2 · Physion overall (frame step 5)accuracy (%): 69.2 -> 71.6Improved
add object markers to the videos (ablation)V-JEPA2 · Physion containmentaccuracy (%): 68.7 -> 78.7Improved
add object markers to the videos (ablation)V-JEPA2 · Physion collisionaccuracy (%): 71.3 -> 78.7Improved
add object markers to the videos (ablation)V-JEPA2 · Physion roll/slideaccuracy (%): 84.0 -> 80.0Hurt
use token-level probe instead of mean-pooled attentive probe (ablation)V-JEPA2 · Physion (frame step 5)accuracy (%): 62.50% -> 67.42%Improved
preserve spatial tokens (CrossAttn-GRU) instead of spatial mean pooling (GRU-Pool) (ablation)V-JEPA2 probe · Physion roll/slideaccuracy (%): +15.3%Improved
use V-JEPA2 backbone instead of VideoMAEv2ViT-L probe · Physionaccuracy (%): 67.65 -> 73.16Improved
use V-JEPA2 backbone instead of DINOv2ViT-L probe · Physionaccuracy (%): 66.90 -> 73.16Improved
require future prediction (target not yet in contact at cutoff)V-JEPA2 · Physion model-human disagreementdisagreement (%): 10.5% -> 26.8%Hurt
majority-vote ensemble of all four models instead of V-JEPA2 alone (ablation)V-JEPA2, RVM, VideoMAEv2, DINOv2 · Physion (10 seeds)accuracy (%): 73.3 -> 71.8Hurt
select probe head by test oracle instead of train accuracy (ablation)V-JEPA2 · Physionaccuracy (pp): 0.36ppNo clear effect
use a 5.75 frame step (1.5s) instead of 5 (1.17s) (ablation)V-JEPA2 · Physion overallaccuracy (%): 71.6 -> 73.2Improved

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