World action models (WAMs) couple generative video modeling with robotic action prediction but typically require large video and action transformers, limiting deployment efficiency. We present LoopedWAM, a compact WAM that recurrently applies a shared six-block core in each expert, executing 30 transformer layers while storing only 12 blocks per expert. To account for their asymmetric inference costs, the video and action experts can operate at different depths: visual features are computed once per query, whereas action computation repeats at every sampling step. We further propose elastic self-distillation, which trains a single elastic checkpoint, LoopedWAM-E, across action-depth and sampling-step budgets. With only 0.58B policy parameters and no robot pretraining, LoopedWAM achieves 96.1% success on LIBERO, 91.9% on RoboTwin 2.0, and 52.2% on RoboCasaGR1, while using 7.4× less peak GPU memory than Fast-WAM under matched inference measurements. On LIBERO, LoopedWAM-E achieves 96.0% success in 27.6 ms or 95.2% in 24.7 ms on an H100, providing multiple accuracy-latency operating points from a single checkpoint; with two sampling steps and four action loops, it stays within 0.5 points of LoopedWAM on all three benchmarks. Ablations demonstrate that weight-shared depth preserves competitive performance, while aggressive action-depth reduction requires explicit budget-aware training. On a real AgileX PiPER arm, LoopedWAM succeeds in 71.3% of 80 trials across four tasks.
Figures & tables
Figure 1: LoopedWAM is small, fast, and elastic. (a) Each expert stores 12 blocks and loops its shared 6-block core K times. One checkpoint thus reaches the 30-layer depth of its dense donor with 2.4× fewer parameters, and the depth is chosen at test time. (b) RoboTwin 2.0 success against parameters, as reported by each paper. At 0.58B, LoopedWAM (91.9%) is the best model below 2B and on par with the ∼ 6B Fast-WAM (91.8%). (c) LIBERO success against latency on one H100 for each setting of action loops Ka and sampling steps N of LoopedWAM-E. One checkpoint spans 24.7–86.2 ms at 95.2–96.2% success, and at (2,2) it is 3.9× faster than Fast-WAM.
Figure 2: Method overview. (a) Each expert stores a prelude, a shared core, and a coda, and loops the core Kv times in the video expert and Ka times in the action expert (shown for Kv=4 , Ka=2 ). The video expert encodes the current frame once per query and caches its keys and values. At every sampling step, the action expert reads them in a joint attention; late alignment lets action loop r attend to video loop σ(r) , so even a single action loop sees the deepest visual features. Both experts also cross-attend to frozen text embeddings of the instruction, and future frames are predicted only during training. (b) One student learns every setting (Ka,N) in S . The target of each macro step is the frozen teacher’s fine Euler map started from the student’s own state ( Llocal ), and the final chunk is matched to the teacher’s chunk from the same noise ( Lend ).
Method
Params
Robot PT
Spatial
Object
Goal
Long
Avg.
OpenVLA-OFT [ 28 ]
7B
✓
97.6
98.4
97.9
94.5
97.1
π0 [ 7 ]
3.3B
✓
96.8
98.8
95.8
85.2
94.2
π0.5 [ 26 ]
3B
✓
98.8
98.2
98.0
92.4
96.9
GR00T-N1.6 [ 6 ]
3.3B
✓
97.7
98.5
97.5
94.4
97.0
SmolVLA [ 49 ]
0.45B
✗
90.0
96.0
92.0
71.0
87.3
RD-VLA [ 52 ]
0.5B
–
92.0
99.0
96.0
84.8
93.0
Table 1: LIBERO success rate (%). Results of prior methods are taken from their papers. ∗ Re-evaluated by us with the released checkpoints under our protocol (40 tasks × 50 initial states × 3 seeds, 6,000 episodes), which we also use for our models. Params: policy parameters without the frozen VAE and text encoder. Robot PT: pretraining on robot data; “–”: not stated. LoopedWAM-E is one checkpoint run with Ka action loops and N sampling steps.
Table 4
Method
Params
Ka
N
Mem.
Latency
Speed-
LIBERO
RoboTwin
(GB)
(ms)
up
(%)
(%)
Prior world action models
Fast-WAM ∗
∼ 6B
–
10
15.25
107.5
1.0 ×
96.6
91.8
Fast-WAM ∗
∼ 6B
–
2
15.32
37.1
2.9 ×
96.0
90.1
Faster-WAM ∗
∼ 6.3B
–
10
12.76
72.6
1.5 ×
98.7
92.6
Light-WAM ∗
1.9B
–
–
4.38
21.8
4.9 ×
95.8
76.4
Table 4: Inference cost in one shared harness (one H100, batch 1, mean over 500 queries). Mem.: peak GPU memory (VAE included, text encoder not); for -E rows it includes memory retained from previously timed settings (see text). Speed-up: relative to Fast-WAM ∗ with its native 10 steps. Rows without “-E” run the undistilled model with N sampling steps; “-E” denotes elastic self-distillation ( Section 3.5 ), applied to Dense-30 over N∈{2,5,10} only. LIBERO: protocol of Table 1 . RoboTwin 2.0: average of clean and randomized ( Table 2 ; prior methods as reported). Light-WAM predicts the action chunk in one forward pass. ∗ Re-evaluated by us.
Figure 3: Elastic inference (LIBERO, shorter ablation schedule; latency: one H100, batch 1). (a) Running fewer action loops than trained collapses the success of the fixed Kv/Ka=4/4 model (orange), and more loops lower it as well, whereas the elastic student keeps it at every loop count. Filled markers: loop counts used in training; hollow: not used (at Ka=4 both models score 94.9%). (b) The nine settings of one elastic checkpoint (blue, Pareto frontier as a line) against models trained separately at fixed loop counts Kv/Ka (the 4/4 model in orange; all fixed models use N=10 ), the 12- and 30-layer dense controls, and a specialist distilled only for (Ka,N)=(1,2) . (c) Latency grows linearly with the number of executed action blocks; the intercept is the per-query cost of encoding, the video expert, and I/O.
Ka
N
Calls
Latency (ms)
Mem. (GB)
Spatial
Object
Goal
Long
Avg.
Gain
1
2
24
24.7
2.58
96.87
99.07
96.60
88.20
95.18
+0.85
2
2
36
27.6
2.68
97.80
99.13
97.80
89.20
95.98
+1.45
4
2
60
33.2
2.78
97.20
99.27
97.53
88.53
95.63
+0.67
1
5
60
33.3
2.62
97.40
98.80
95.80
88.93
95.23
+0.70
2
5
90
39.1
2.72
97.67
98.93
96.87
88.73
95.55
+0.70
1
10
120
47.9
2.65
97.27
99.27
96.07
88.73
95.33
+1.65
Table 5: All nine operating points of LoopedWAM-E on LIBERO (success %, 6,000 episodes each). Calls: action-block executions per query, N(6+6Ka) . Latency and peak memory: one H100, batch 1. Shaded: the two settings reported in the main tables. Keeping all nine settings loaded at once needs 3.18 GB. Gain: change from the student distilled from the shorter-schedule model at the same setting ( Table B1 ), paired over the same 6,000 episodes; significant at the 5% level in seven of the nine settings (permutation test, unadjusted).
RoboTwin 2.0
RoboCasa-GR1
Ka
N
Calls
Clean
Rand.
Avg.
Avg.
1
2
24
89.5
89.0
89.3
43.2
2
2
36
90.6
90.6
90.6
51.6
4
2
60
91.3
91.6
91.5
52.4
1
5
60
89.3
89.6
89.4
47.4
2
5
90
90.6
90.9
90.7
50.8
Table 6: LoopedWAM-E on RoboTwin 2.0 and RoboCasa-GR1 : all nine settings of each benchmark’s student (success %; RoboTwin 2.0: 100 episodes per task in each setting; RoboCasa-GR1: 50 rollouts per task). Calls: action-block executions per query. Shaded: the settings of the main tables.
(a) Depth and loop allocation
Model
Kv
Ka
Params
LIBERO
RT-10
Dense-12
–
–
0.58B
90.0
77.3
LoopedWAM
1
4
0.58B
91.7
79.6
LoopedWAM
2
2
0.58B
92.5
78.9
LoopedWAM
4
1
0.58B
94.0
81.0
LoopedWAM
4
2
0.58B
94.5
82.4
Table 7: Ablations on the shorter schedule (success, %; LIBERO: 6,000 episodes per model; RoboTwin-10 (RT-10): 2,000 episodes per model). Kv/Ka : video and action loops; Δ : change from the default (shaded). † Trained for half as many epochs as the other RT-10 models. Per suite: Appendix C .
LIBERO
RoboTwin-10
Objective
(2,2)
(4,2)
(1,10)
Avg.
Δ
(2,2)
(4,2)
(1,10)
Avg.
Δ
LoopedWAM-E ( Eq. 6 )
95.98
95.63
95.33
95.65
–
80.30
81.25
75.30
78.95
–
Teacher trajectory
95.26
95.98
95.08
95.44
−0.21
78.95
80.55
73.75
77.75
−1.20
w/o Lanchor
95.58
95.50
94.88
95.32
−0.33
78.70
80.75
75.40
78.28
−0.67
w/o Lend
95.93
95.65
94.78
95.45
−0.20
78.50
81.10
76.05
78.55
−0.40
w/o Lgrip
95.88
95.38
95.25
95.50
−0.15
n/a (continuous grippers)
Table 8: Distillation objective (success %). Each row distills LoopedWAM-E again with one change to Eq. 6 and otherwise the same teacher, initialization, data order, noise, and sequence of settings. Teacher trajectory : every student step, and the local target that supervises it, starts from the teacher’s state at τj instead of the student’s. LIBERO: students of the main LoopedWAM, 6,000 episodes per setting; Avg. over all nine settings; Δ : change of that average, paired over the same 54,000 episodes. RoboTwin-10: students of the 4/4 model, 2,000 episodes per setting; Avg. over the three settings shown; Δ : unpaired difference; RoboTwin’s grippers are continuous, so its objective has no Lgrip .
Figure 4: Real-world setup and tasks. (a) The follower AgileX PiPER arm with a front and a wrist camera; demonstrations are collected by teleoperation with a second PiPER as the leader arm. (b–e) Start (top) and end (bottom) states of the four tasks. Instructions: “pick up the can and place it on the plate”, “press the {blue, orange, pink, red} button”, “stack the bowls”, “put the spoon in the bowl”.
LoopedWAM-E (Ka,N)
Task
Fast-WAM ∗
LoopedWAM
(4,2)
(2,2)
Place Can on Plate
14/20
14/20
14/20
13/20
Press Color Button
18/20
18/20
17/20
17/20
Stack Bowls
14/20
14/20
14/20
14/20
Put Spoon in Bowl
11/20
11/20
10/20
10/20
Total
57/80
57/80
55/80
54/80
Table 9: Real-robot results (successful trials out of 20 per task and method, randomized object poses). All models are trained on the same 100 demonstrations per task. ∗ Fine-tuned and evaluated by us. Latency is measured on the robot computer (RTX PRO 6000).
Figure 5: Successful real-robot rollouts of LoopedWAM (front camera; frames left to right in time). From top: placing the can on the plate; one button board under all four instructions, shown as the shared start frame and the final frame for each instruction; stacking the bowls; putting the spoon in the bowl.
Figure 6: Typical failures of LoopedWAM (front camera; frames left to right in time). Missed grasp: the fingers close beside the can, and the arm moves on to the plate without it. Knocked over: the can tips over as the fingers close and rolls away. Pushed away: the carried bowl is lowered onto the rim of the other bowl and pushes it out of view. Released off target: the gripper opens above the bowl, but the spoon stays hooked on one finger.
Appendix figures & tables10 assets
Supplementary material from the paper’s appendix.
Appendix
Method
Reported
Ours
Light-WAM [ 35 ]
97.2
95.77
Fast-WAM [ 63 ]
97.6
96.63
Faster-WAM [ 64 ]
99.0
98.72
Appendix
Table A1: Re-evaluated LIBERO baselines : average success (%) reported by each paper and obtained under our protocol with the released checkpoints (6,000 episodes).
Ka
N
Lat. (ms)
Spatial
Object
Goal
Long
Avg.
1
2
24.7
96.47
98.93
95.27
86.67
94.33
2
2
27.6
96.40
99.27
95.67
86.80
94.53
4
2
33.2
97.00
98.80
96.33
87.73
94.97
1
5
33.3
97.07
99.07
96.53
85.47
94.53
2
5
39.1
96.47
99.00
96.13
87.80
94.85
1
10
47.9
96.73
98.60
93.87
85.53
93.68
Appendix
Table B1: Elastic student on the shorter schedule (LIBERO success %), the specialist distilled only for (1,2) , and the teacher; data of Figure 3 b.
Ka
Spatial
Object
Goal
Long
Avg.
Fixed 4/4 model
1
18.73
25.13
5.07
0.27
12.30
2
65.07
72.13
54.60
30.47
55.57
3
95.53
92.60
92.93
76.80
89.47
4
96.93
98.73
96.13
87.87
94.92
5
94.47
98.87
96.20
88.87
94.60
Appendix
Table B2: Running with Ka action loops at test time ( N=10 , shorter schedule): the fixed 4/4 model (trained with Ka=4 : 94.92%) and LoopedWAM-E at loop counts it was not trained with; data of Figure 3 a. LoopedWAM-E at its trained loop counts is in Table B1 .
LIBERO
Model
Kv
Ka
Params
Lat. (ms)
Spatial
Object
Goal
Long
Avg.
RT-10
Dense-12
–
–
0.58B
40.8
94.80
93.27
92.13
79.73
89.98
77.25
LoopedWAM
1
4
0.58B
86.4
96.00
97.80
94.73
78.20
91.68
79.55
LoopedWAM
2
2
0.58B
58.3
96.93
96.53
94.27
82.07
92.45
78.90
LoopedWAM
4
1
0.58B
48.3
96.27
98.07
95.40
86.27
94.00
80.95
LoopedWAM, KV-mix
4
1
0.58B
50.9
97.40
98.33
96.93
85.13
94.45
–
Appendix
Table C1: Loop allocation and dense controls on the shorter schedule (LIBERO per suite and RoboTwin-10). Latency: one H100, batch 1. KV-mix lets the single action loop attend to a mixture of all video loops. † Trained for half as many epochs as the other RT-10 models.
Core
Spatial
Object
Goal
Long
Avg.
Blocks 10–15 (default)
96.93
98.73
96.13
87.87
94.92
Blocks 16–21
96.20
97.27
94.60
84.40
93.12
Every 4th block (5, 9, …, 25)
94.93
95.73
93.87
81.33
91.47
Mean of 4 blocks per slot
96.27
96.47
94.40
84.13
92.82
FFN units by weight magnitude
95.80
98.27
92.93
81.47
92.12
Random core
92.00
91.07
85.13
60.60
82.20
Appendix
Table C2: Initialization of the video core (LIBERO success %, shorter schedule). Prelude and coda copy Wan2.1 blocks 1–3 and 28–30.
P/C/E
K
Donor blocks
Spatial
Object
Goal
Long
Avg.
3/6/3
4
1–3 / 10–15 / 28–30
96.93
98.73
96.13
87.87
94.92
0/6/6
4
– / 13–18 / 25–30
97.20
98.73
95.27
86.47
94.42
1/6/5
4
1 / 13–18 / 26–30
97.27
98.13
95.93
83.33
93.67
5/6/1
4
1–5 / 13–18 / 30
96.73
97.40
93.60
84.93
93.17
6/6/0
4
1–6 / 13–18 / –
95.07
92.67
91.27
82.07
90.27
1/9/2
3
1 / 11–19 / 29–30
97.53
98.33
93.47
77.80
91.78
Appendix
Table C3: Partition P/C/E with K loops (LIBERO success %, shorter schedule); donor blocks given as prelude / core / coda.
Task
Dense-12
1/4
2/2
4/1
4/2
4/4
Dense-30 †
adjust_bottle
99.5
99.5
99.5
99.0
100.0
99.5
99.5
grab_roller
100.0
100.0
100.0
100.0
100.0
100.0
99.5
stack_blocks_two
98.0
97.5
98.0
99.0
97.0
97.5
100.0
press_stapler
90.5
88.5
91.0
84.5
85.0
92.5
91.5
stamp_seal
87.0
83.0
90.0
89.0
91.5
89.5
93.5
scan_object
79.5
83.0
87.0
83.0
85.5
84.5
91.0
Appendix
Table D1: RoboTwin-10 per task (success %, average of clean and randomized, 200 episodes per task) for each configuration ( Kv/Ka ). † Trained for half as many epochs as the other models.
Method
Roll.
PnP + close
Cutting board
Place- mat
Tray
Plate
Avg.
GR00T-N1.6 [ 6 ]
–
24.2
56.9
51.9
55.1
57.6
47.6
StarVLA [ 13 ]
50
50.3
52.8
38.0
39.2
58.5
47.8
DiT4DiT [ 44 ]
50
50.3
57.6
39.0
46.4
60.0
50.8
ABot-M0 [ 59 ]
50
61.7
61.6
51.0
56.8
58.0
58.3
LoopedWAM
50
52.7
55.2
46.5
48.8
57.5
52.2
LoopedWAM-E (4, 2)
50
53.0
61.2
46.0
46.0
55.0
52.4
Appendix
Table E1: RoboCasa-GR1 by task group (success %; number of tasks in parentheses in Table E3 ). Roll.: rollouts per task (–: not stated). Prior methods: means of the per-task results in their papers (GR00T-N1.6: as listed by StarVLA); LoopedWAM-E: (Ka,N) .
Task
Clean
Rand.
Avg.
Task
Clean
Rand.
Avg.
adjust_bottle
100
99
99.5
place_can_basket
61
57
59.0
beat_block_hammer
100
96
98.0
place_cans_plasticbox
99
96
97.5
blocks_ranking_rgb
99
99
99.0
place_container_plate
96
97
96.5
blocks_ranking_size
98
96
97.0
place_dual_shoes
91
89
90.0
click_alarmclock
99
100
99.5
place_empty_cup
100
100
100.0
click_bell
100
100
100.0
place_fan
96
93
94.5
Appendix
Table E2: RoboTwin 2.0 per task for LoopedWAM (successes out of 100 in the clean and randomized settings, and their average in %).
Task
LoopedWAM
LoopedWAM-E (4,2)
LoopedWAM-E (2,2)
Bottle → cabinet, close
58.0
62.0
58.0
Can → drawer, close
68.0
76.0
70.0
Cup → drawer, close
44.0
34.0
48.0
Milk → microwave, close
48.0
50.0
46.0
Potato → microwave, close
44.0
42.0
36.0
Wine → cabinet, close
54.0
54.0
48.0
Appendix
Table E3: RoboCasa-GR1 per task for LoopedWAM and LoopedWAM-E at (4,2) and (2,2) (success %, 50 rollouts per task).