RealtimeWAM: One-Step Asynchronous World Action Models

World Action Models (WAMs) incorporate visual representations from video generation backbones to guide action prediction.

By Lv, Du, Feng +7Oct 5, 2026

Score█████░░░░░5.5

Key numbers

1 % drop

VerdictWorth a reader's time today.

Read the original PDF Code

Abstract

World Action Models (WAMs) incorporate visual representations from video generation backbones to guide action prediction. Recent efficient WAMs adopt Mixture-of-Transformers (MoT) architectures and compute video representations once for reuse by the action expert. However, intra-expert iteration (\ie, multi-step action denoising) and inter-expert waiting (\ie, sequential execution of the video and action experts) still limit inference efficiency. To this end, we present RealtimeWAM, an extremely efficient WAM variant with one-step action generation and asynchronous inference, addressing these two bottlenecks. To reduce intra-expert iteration, we propose Teacher-Anchored Consistency Distillation (TACD) to address a local-global error gap: low local consistency error alone does not guarantee accurate final actions. TACD supplements local consistency with explicit supervision from the frozen teacher's multi-step rollout endpoint, enabling accurate one-step action generation. Additionally, we propose Cross-Expert Wavefront Pipelining (CEWP) to eliminate unnecessary expert-level waiting. It overlaps the two experts through block-wise sharing of the video KV cache, synchronizing only immediately before the corresponding action attention consumes it. Extensive experiments across diverse benchmarks (\eg, LIBERO, LIBERO-Plus and RoboTwin) and model variants (\eg, Fast-WAM and Faster-WAM) demonstrate the superiority of RealtimeWAM. Notably, RealtimeWAM maintains near-lossless performance (\ie, <1% drop) across these benchmarks while delivering significant end-to-end speedup (\eg, ~25× on H100). Our code and checkpoints are available via this \href{https://github.com/ModelTC/LightX2V/tree/main/examples/realtimewam}{link}.

Chengtao Lv, Jinyang Du, Shuyi Feng, Yang Yong, Shiqiao Gu, Shunzi Yang, Ruihao Gong, Shen Ren, Tianwei Zhang, Wenya Wang

The editor's rubric

Heuristic review

Dimension	Level	Weight	What that level means
Leverage	████░ 4	24%	A general-purpose tool used across several fields (Adam, ResNet, LoRA, next-generation sequencing).
Magnitude	██░░░ 2	18%	Solid incremental gain on a meaningful problem.
Evidence	███░░ 3	14%	Solid: multiple benchmarks or cohorts, ablations, fair baselines, released code or data.
Novelty	██░░░ 2	16%	A new combination of known ideas.
Trajectory	███░░ 3	18%	A clear path to scale.
Stakes	██░░░ 2	10%	Benefits a professional community (practitioners, clinicians, engineers).

Editor’s rationale

Heuristic triage from title and abstract text only, not a reading of the paper. Cues found: method (we propose); breadth (many tasks); verification (code released); scale (efficient); stakes (global scale).

How the score was computed

rank-2026-09-29

Score█████░░░░░5.5

Score = 10 × (65% × adjusted merit / 10 + 25% × attention + 10% × freshness)

Merit: 5.6 / 10; Weighted rubric, evidence-gated.
Adjusted merit: 4.6 / 10; Shrunk toward the desk prior by editor confidence (38%).
Attention: 69%; Citations, upvotes, points, mentions.
Freshness: 78%; Half-life decay since publication.

Hugging Face upvotes7 (reference 25, via hf-daily, Oct 6, 2026, 13:49 UTC)
GitHub stars2.9k (reference 250, via hf-daily, Oct 6, 2026, 13:49 UTC)

The record

Reviewed by heuristic-v2 on Oct 6, 2026, 13:49 UTC. Paper type: method.
Categories: cs.CV, cs.LG, cs.RO
TOP, No.3 in the Artificial Intelligence edition of October 6, 2026.