MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling

Jie Chen; Ruofei Bai; Yuxin Cai; Yifeng Zhang; Chengyang He; Jun Li; Wei-Yun Yau; Guillaume Sartoretti

cs.ROOct 8, 2026

MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling

Authors: Jie Chen, Ruofei Bai, Yuxin Cai, Yifeng Zhang, Chengyang He, Jun Li, Wei-Yun Yau, Guillaume Sartoretti

Organizations: Department of Mechanical Engineering, National University of Singapore, Singapore · A*STAR Institute of Advanced Intelligence and Computing (A*STAR IAIC), Singapore · Nanyang Technological University, Singapore

Abstract

World modeling has emerged as an effective co-training objective for robot policies, giving rise to World Action Models (WAMs) that jointly predict actions and future states. However, most WAMs predict future states in the native representation space of pretrained visual backbones, resulting in high-dimensional targets with substantial training cost. We introduce MiniWAM, which instead predicts compact future representations learned from privileged current-future transitions. To construct these targets, we propose Predictive Representations via Inverse Spatiotemporal Modeling (PRISM), which combines inverse-dynamics supervision with feature reconstruction to emphasize control-relevant transition information while preserving useful future-state information. With the learned PRISM encoder frozen, MiniWAM is trained to jointly predict the resulting targets and robot actions from current observations. With 65 $\times$ fewer native future feature tokens, MiniWAM consistently outperforms native future-feature prediction with both DINOv3 and WAN2.1 VAE features, while achieving up to an 8 $\times$ speedup in world-action training. At 0.25B parameters, MiniWAM is already competitive with substantially larger WAMs on LIBERO, LIBERO-Plus, and RoboTwin 2.0 simulation benchmarks. Representation analyses further show that PRISM contributes behavioral structure beyond feature reconstruction alone. These results demonstrate that effective world-action modeling does not require predicting native visual futures, and that compact predictive representations provide a strong and substantially more efficient target for policy learning. The project page is available at: https://j1dan.github.io/MiniWAM.

Figures & tables

Figure 1: MiniWAM replaces costly native future prediction with compact predictive representations. (a) While conventional WAMs predict high-dimensional native future features, MiniWAM jointly predicts actions and compact PRISM targets learned from the same visual features via inverse dynamics and reconstruction. (b) MiniWAM remains competitive with substantially larger WAMs and VLAs on LIBERO-Plus . (c) In controlled comparisons across VAE and DINOv3 feature spaces, PRISM simultaneously improves success rate and reduces training time relative to native future-feature prediction.

Figure 2: Two-stage training pipeline of PRISM and MiniWAM. In Stage 1 , privileged current–future visual features are compressed by the PRISM encoder into a compact predictive representation $L_{t}$ , shaped through inverse action prediction and feature reconstruction. The corresponding world tokens remain clean while the action tokens are corrupted for conditional action prediction. In Stage 2 , the PRISM encoder is frozen and provides compact world-modeling targets; MiniWAM jointly denoises the world and action tokens using its world and action experts. Solid and hatched tokens denote clean and noised tokens, respectively. Language and proprioceptive conditioning are omitted for clarity.

Visual features	Co-training target	LIBERO	LIBERO-Plus-2K
		SR (%) $\uparrow$	SR (%) $\uparrow$
VAE	w/o Co-Train	62.3	25.1
	w/ Native Co-Train	92.8	50.2
	w/ PRISM	94.8	56.5
DINOv3	w/o Co-Train	89.0	56.5
	w/ Native Co-Train	93.2	60.9

Table 1: Controlled comparison of world-modeling targets on LIBERO and LIBERO-Plus-2K . All variants use the same 0.25B world–action architecture and training protocol within each visual feature family.

		Horizon 16			Horizon 32
Visual Feature	Target	Throughput	S2	Total	Throughput	S2	Total
		$\uparrow$	$\uparrow$	$\uparrow$	$\uparrow$	$\uparrow$	$\uparrow$
LIBERO
WAN2.1 VAE	Native	30.81	1.00 $\times$	1.00 $\times$	14.20	1.00 $\times$	1.00 $\times$
	PRISM	142.04	4.61 $\times$	3.55 $\times$	124.92	8.80 $\times$	6.77 $\times$
DINOv3	Native	36.17	1.00 $\times$	1.00 $\times$	15.55	1.00 $\times$	1.00 $\times$

Table 2: Training efficiency of PRISM versus native co-training. Throughput is measured in samples/s with global batch 256 under matched hardware. S2 is Stage-2 speedup; Total estimates training speedup including Stage 1, assuming equal per-update costs for the two PRISM stages (Appendix D ).

Type	Method	Scale	Embodied PT.	Backbone Init.	LIBERO SR (%) $\uparrow$	LIBERO-Plus SR (%) $\uparrow$	RoboTwin 2.0 SR (%) $\uparrow$
VLA	OpenVLA-OFT	7B	$\checkmark$	VLM	96.6	71.4	-
	$\pi_{0}$	3.3B	$\checkmark$	VLM	93.5	70.5	31.4
	$\pi_{0}$ -FAST †	3.3B	$\checkmark$	VLM	85.5	61.6	-
	UniVLA †	7B	$\checkmark$	VLM	95.2	42.9	-
	StarVLA-OFT †	4B	$\times$	VLM	96.6	-	24.8
	Xiaomi-Robotics-0 †	4.7B	$\checkmark$	VLM	98.7	-	40.5

Table 3: Comparison with state-of-the-art robot policies.

\dagger

denotes externally reported results. Scale is the reported model size; Embodied PT and Backbone Init. denote embodied pretraining and backbone initialization. “–” means no pretrained backbone and “-” means unavailable.

Bottleneck	Recon.	Inv. dyn.	Both
$4{\times}4$ , $D{=}64$	69.5	70.5	74.1
$2{\times}2$ , $D{=}64$	63.3	67.5	67.8
$1{\times}1$ , $D{=}16$	57.0	56.8	61.7

Table 4: PRISM ablation on LIBERO-Plus-2K . Success rate (%).

Protocol	Raw	Random	PRISM
Global	.888/.571	.891/.469	.889/ .736
Within-task	.372/.249	.682/.347	.795 / .705
Deconfounded	.109/.071	.162/.162	.379 / .393

Table 5: Cross-backbone alignment on unseen LIBERO-90 tasks. Entries report main/wrist-view CKA.

Metric	Main	Wrist
Jaccard@5 $\uparrow$	0.170 $\rightarrow$ 0.512	0.107 $\rightarrow$ 0.333
Action RMSE $\downarrow$	0.400 $\rightarrow$ 0.365	0.370 $\rightarrow$ 0.272
EEF error (m) $\downarrow$	0.083 $\rightarrow$ 0.072	0.089 $\rightarrow$ 0.076
Gripper agr. $\uparrow$	0.758 $\rightarrow$ 0.793	0.820 $\rightarrow$ 0.957

Table 6: Behavioral structure on unseen LIBERO-90 tasks. Leave-task-out retrieval results are reported as Raw

\rightarrow

PRISM.

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

Setting	Stage 1	Stage 2
Optimizer; peak LR; decay	AdamW; $5\cdot 10^{-4}$ ; 0	same
LR schedule; warmup	cosine; 5% of updates	same
Global batch; precision; seed	256; BF16; 42	same
LIBERO duration	15 epochs	50 epochs
RoboTwin duration	6 epochs	25 epochs
LIBERO action / visual horizon	16 / 16	same

Appendix

Table 7: Shared 0.25B training settings. Microbatch and gradient accumulation vary with GPU count while their product with GPU count remains 256.

Method	LIBERO	LIBERO-Plus	RoboTwin 2.0
$\pi_{0}$ -FAST	85.5 ( Fei et al., 2026 )	61.6 ( Fei et al., 2026 )	–
UniVLA	95.2 ( Bu et al., 2025 )	42.9 ( Fei et al., 2026 )	–
StarVLA-OFT	96.6 ( Community, 2026 )	–	24.8 ( RoboTwin Team, 2026 )
Xiaomi-Robotics-0	98.7 ( Cai et al., 2026b )	–	40.5 ( RoboTwin Team, 2026 )
WorldVLA	79.1 ( Cen et al., 2025 ; Fei et al., 2026 )	25.0 ( Fei et al., 2026 )	–
AHA-WAM	–	–	33.8 ( RoboTwin Team, 2026 )

Appendix

Table 8: Sources of externally reported results in Table 3 . Values are success rates (%).

Benchmark	Visual	Native		PRISM		Pipeline
	horizon	Core	Pipeline	Core	Pipeline	speedup
LIBERO	16	36.17	35.16	143.25	138.81	$3.95\times$
	32	15.55	15.26	126.09	123.08	$8.07\times$
RoboTwin 2.0	16	21.48	21.11	93.57	92.86	$4.40\times$
	32	8.61	8.41	81.47	80.74	$9.60\times$

Appendix

Table 9: DINOv3 training throughput through the data pipeline. Samples per second on one RTX PRO 6000 GPU at global batch 256. Core values reproduce the compute measurements in Table 2 ; pipeline values include data loading and transfer. Speedup compares PRISM and native Stage-2 pipeline throughput.

View	Objective	Action RMSE $\downarrow$	EEF Err. (m) $\downarrow$	Grip. Agr. $\uparrow$
Main	Reconstruction only	0.414	0.090	0.747
	PRISM (I.D. + recon.)	0.356	0.068	0.802
Wrist	Reconstruction only	0.326	0.087	0.907
	PRISM (I.D. + recon.)	0.266	0.078	0.975

Appendix

Table 10: Effect of inverse-dynamics supervision on neighborhood behavior. Matched DINOv3 bottlenecks on unseen LIBERO-90 tasks; both retain the full posterior-mean grid.

Explore similar work

Light-WAM: Efficient World Action Models with State-Fusion Action Decoding
Jun 6, 2026Ziang Li, Dongzhou Cheng, Yibin Wang +5Robot Policy Learning World Models for Robotics
Wuhan University · Shanghai Innovation Institute · Southeast University +2
MoWAM: Explicit Future Motion Prediction for Efficient World Action Models
Sep 17, 2026Jiayu Wang, Bin Zhu, Yue Yu +1World Models for Robotics World Model-Based Planning
College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, China · Singapore Management University, Singapore · Institute of Trustworthy Embodied AI, Fudan University, Shanghai, China
Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination
Jun 8, 2026Jiajun Li, Tiecheng Guo, Yifan Ye +9World Models for Robotics Efficient Inference for World Action Models
1The University of Hong Kong · 2Peking University · 3Muka Robotics +2

Jun 6, 2026, cs.CVJ/K move · Enter open · S save

Light-WAM: Efficient World Action Models with State-Fusion Action Decoding

Ziang Li, Dongzhou Cheng, Yibin Wang, Shiyue Wang+4

Wuhan University · Shanghai Innovation Institute · Southeast University+2

Robot Policy Learning World Models for Robotics Visuomotor Policy Learning Efficient Inference for World Action Models World Action Models

Open paper AskPDF arXiv

World Action Models (WAMs) extend robot policy learning by incorporating future prediction as an additional training objective, encouraging the policy to encode task-relevant temporal structure in its representations. Current WAMs often rely on large-scale generative architectures that incur high training costs and inference latency, making them difficult to deploy as efficient closed-loop policies. We propose Light-WAM, a lightweight World Action Model for efficient robot manipulation. Specifically, it is built with a compact video backbone and performs future-video supervision in a downsampled latent space, reducing the cost of video co-training while retaining its benefits for representation learning. For action prediction, Light-WAM introduces the StateFusionActionExpert, which reads adapted states from multiple backbone layers, fuses them through learned-query pooling, and directly predicts action chunks in a single forward pass. This design provides an efficient interface between video backbone representations and robot actions, avoiding the need for heavy generative action experts. Experiments demonstrate that Light-WAM maintains strong performance on LIBERO and achieves usable multi-task performance on RoboTwin 2.0, while using only 0.44B trainable parameters. It also achieves 72.03ms inference latency with 4.1GiB peak GPU memory and improved training throughput.

Abstract

Explore similar work

Light-WAM: Efficient World Action Models with State-Fusion Action Decoding

MoWAM: Explicit Future Motion Prediction for Efficient World Action Models

Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination