AI Paper Atlas

Research map / Video / Action recognition

SwiftWA: An Efficient Action-Centered World-Action Model

Chaojun Ni, Xinyu Zhou, YuKun Zhou, Jingyu Liu, Xiaofeng Wang, Zheng Zhu, Yang Wang, Qiuping Deng, Yun Ye, Hao Li, Zhichao Liu, Jindi Lv and 5 more

ECCV 2026 · Video › Action recognition · approach: Action anticipation and latent actions

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 future-video prediction supervision (stride 12) during training (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 0.60 -> 0.83Improved
predict too many (stride 4) or too few (stride 48) future frames instead of stride 12 (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 0.83 -> 0.76Hurt
add embodied data pre-training on top of video-model initialization (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 0.57 -> 0.83Improved
initialize from pretrained video model instead of scratch (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 0.45 -> 0.57Improved
increase embodied pre-training data fraction (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 57% -> 83%Improved
use causal attention mask instead of full self-attention (ablation)GigaWorld-Policy · Real-world taskssuccess rate: 0.81 -> 0.83No clear effect
use causal attention mask instead of full self-attention (ablation)GigaWorld-Policy · Real-world held-out videoPSNR: 27.87 -> 28.41Improved
decode actions only at inference (skip video rollout) vs MotusGigaWorld-Policy · Inference latencylatency (ms): 3231 -> 360Improved
use GigaWorld-Policy instead of MotusGigaWorld-Policy · RoboTwin 2.0 averagesuccess rate: 0.89 -> 0.87 (clean)Hurt
use GigaWorld-Policy instead of MotusGigaWorld-Policy · Real-world taskssuccess rate: 0.76 -> 0.83Improved
use GigaWorld-Policy instead of pi0.5GigaWorld-Policy · RoboTwin 2.0 average (clean)success rate: 0.43 -> 0.87Improved
use GigaWorld-Policy instead of MotusGigaWorld-Policy · Real-world OODsuccess rate: 0.60 -> 0.72Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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