Completion Aware Guidance for World Action Models

Seungyeon Kim; Junhoo Lee; Baekseung Kim; Minkyu Kim; Nojun Kwak

cs.ROOct 1, 2026

Completion Aware Guidance for World Action Models

Authors: Seungyeon Kim, Junhoo Lee, Baekseung Kim, Minkyu Kim, Nojun Kwak

Organizations: Seoul National University · KAIST

Abstract

World Action Models (WAMs) predict visual futures and robot actions, yet they remain susceptible to task-incomplete imagination, where plausible, action-consistent predictions omit the transition needed for task completion. In this paper, we show that this failure is not inherent to the world model backbone, but emerges when adapted for short-chunk control, which can repeatedly favor plausible local continuations over task-completing transitions. To address this, we introduce Completion Aware Guidance (CAG), a training-free sampling method that guides generation toward task completion. Across representative WAMs, CAG improves success from 64% to 70% on a RoboTwin 2.0 subset and from 69% to 75% in zero-shot simulation, while reducing task-incomplete imagination from 79% to 40%.

Figures & tables

Figure 1: Completion Aware Guidance. A standard WAM produces a visually plausible, action-consistent continuation while omitting the transition required for task completion. CAG strengthens instruction conditioning relevant to the current prediction and recovers the transition during control.

Panel of Figure 2 — Figure 2: Analysis of CAG on DreamZero. (a) Video cㅁompletion rates under standard 1.6-second chunk generation and extended 5-second generation, showing that a longer generation horizon recovers task-completing transitions. (b) Zero-shot task-success rates for DreamZero and CAG. (c) Proportion of failed rollouts attributed to task-incomplete imagination, which is reduced by CAG.

Task	$\pi_{0.5}$ [ 5 ]	GO-1 [ 25 ]	Fast-WAM [ 7 ]	CAG
Open Microwave	37	14	33	30
Hanging Mug	3	0	60	63
Turn Switch	6	30	53	67
Place Can Basket	25	37	63	53
Move Stapler Pad	18	4	57	73
Stack Bowls Three	35	7	67	87

Table 1: RoboTwin 2.0 results. CAG improves overall task success on non-saturated tasks.

Explore similar work

Latent Action as Intention Enables Efficient Future Imagination for World Action Models
Aug 25, 2026Xiang Li, Yupeng Zheng, Songen Gu +11Efficient World-Action Model Latent Actions
1CollegeAI, THU · 2AIR, THU · 4TARS Robotics +6
When to Trust Imagination: Adaptive Action Execution for World Action Models
May 7, 2026Rui Wang, Yue Zhang, Canyang Chen +3Efficient World-Action Model World Models
Southern University of Science and Technology, Shenzhen, China · The University of Hong Kong, Hong Kong, China
Making Foresight Actionable: Repurposing Representation Alignment in World Action Models
Jun 10, 2026Lu Qiu, Yizhuo Li, Yi Chen +3Efficient World-Action Model World Models
The University of Hong Kong · XPENG Robotics

Aug 25, 2026, cs.ROJ/K move · Enter open · S save

Latent Action as Intention Enables Efficient Future Imagination for World Action Models

Xiang Li, Yupeng Zheng, Songen Gu, Huailiang Ma+10

1CollegeAI, THU · 2AIR, THU · 4TARS Robotics+6

Efficient World-Action Model Latent Actions Faster-Wam Imagination World Models Robocasa

Open paper AskPDF arXiv

World action models (WAMs) improve robot control by modeling how observations evolve, but generating future observations at test time incurs substantial latency. Fast-WAM removes this process for efficiency; however, our matched implementations show lower generalization for Fast-WAM than for future-aware alternatives, especially with scarce robot demonstrations and in out-of-distribution scenarios. To bridge this gap, we introduce LAWA, a WAM architecture that uses compact latent actions as an operational representation of future intentions, enabling efficient test-time future imagination without generating future observations. Specifically, a discrete tokenizer enhanced by action-free pre-training produces manipulation-centric codebook targets. LAWA jointly denoises a continuous latent state anchored to these targets with executable action chunks while omitting the future-video branch at inference. On RoboCasa, LAWA achieves state-of-the-art average success rates of 65.6% and 80.8% in the few-shot and full data settings, improving over the matched Fast-WAM baseline by 9.6 and 4.5 points, respectively. It also preserves the performance level of the matched Joint-WAM variant while requiring 42.9% lower inference latency. LAWA also demonstrates competitive zero-shot robustness on LIBERO-Plus and superior performance on real-world tasks. These results show that future imagination need not be discarded: retaining it with compact latent actions yields an effective trade-off among performance, generalization, and latency. Code and models will be released.

Abstract

Explore similar work

Latent Action as Intention Enables Efficient Future Imagination for World Action Models

When to Trust Imagination: Adaptive Action Execution for World Action Models

Making Foresight Actionable: Repurposing Representation Alignment in World Action Models