Blackout vs. Freeze: Analyzing Physical Failure Modes of VLAs under Camera Faults

Heejae Suh; Jongwook Han; Zahra Gholami; Yohan Jo

cs.ROSep 30, 2026

Blackout vs. Freeze: Analyzing Physical Failure Modes of VLAs under Camera Faults

Authors: Heejae Suh, Jongwook Han, Zahra Gholami, Yohan Jo

Organizations: Seoul National University, Seoul, Republic of Korea

Abstract

Unreliable visual inputs can harm task performance and cause potential physical safety risks for vision-language-action (VLA) models. We analyze how $π0.5$ and GR00T models act under input faults such as image blackouts and freezing. We find that blackout and freezing produce distinct physical failure modes even when task-success rates are similarly low: freezing causes more extreme joint behavior, whereas blackout after gripper closure can cause more object drops, most markedly without proprioception. Selective intervention studies reveal that proprioception (current robot state) partly compensates for the removed robot depictions and reduces non-target contact. However, it cannot sufficiently restore task success when wrist-view object information is removed, even when aided by the remaining scene view. We then evaluate two mitigation approaches: camera-blackout training and training-free replacement of faulty visual embeddings. Both improve task success in selected conditions, but can increase unintended contact or disturbance to surrounding objects. Real-robot trials further show that successful execution under camera faults can still involve unintended physical interactions. These findings motivate designing VLA policies that use the robot and object information still available under camera faults to limit hazardous motion.

Figures & tables

Figure 1: Overview of the camera-fault evaluation protocol.

Panel of Figure 2 — Figure 2: Task performance and executed commands under camera faults, with and without proprioception. (a) Task success under clean input, blackout, and freezing, pooling 150 episodes from each of Spatial, Object, and Goal ( $n=450$ per condition). (b) Gripper commands under clean, scene-blackout, and wrist-blackout inputs. Each panel displays the same 30 selected episodes, ordered within a policy by the clean first-close command; success labels use all 150 Spatial episodes. Gray denotes open commands and white denotes time after termination.

Measure	Event or statistic	Reported as
Core physical outcomes
NTC	Robot–object contact, excluding target and arena	% of steps
JVE	At least one joint exceeds its velocity limit	% of episodes or steps
JLP	Normalized J2 or J7 position enters the registered 0.05 limit margin	% of episodes
OD	Non-target object displacement $>2$ cm	% of episodes
Additional physical measures

Table 1: Safety-relevant physical measures.

$\pi 0.5$
			Execution context		Contact and object effects				Joint motion
Proprio.	Camera	Fault	SR	$\\|\Delta xyz\\|$	NTC	OD	Drop	$F_{5}$	JVE	JLP
Yes	–	Clean	95 .3 $\to$ 95 .3	0 .76 $\to$ 0 .76	2 .0 $\to$ 2 .1	2 .0 $\to$ 1 .3	1 .3	0 .0 $\to$ 0 .0	0 .0 $\to$ 0 .0	1 .3 $\to$ 1 .3
	Scene	Black	23 .3 $\to$ 66 .0	0 .58 $\to$ 0 .67	7 .2 $\to$ 2 .6	14 .0 $\to$ 6 .7	2 .0	1 .3 $\to$ 2 .0	0 .0 $\to$ 0 .0	2 .0 $\to$ 6 .0
	Scene	Frozen	2 .7 $\to$ 6 .7	0 .71 $\to$ 0 .70	11 .7 $\to$ 3 .4	39 .3 $\to$ 12 .7	5 .3	36 .7 $\to$ 0 .7	85 .3 $\to$ 41 .3	85 .3 $\to$ 42 .7
	Wrist	Black	21 .3 $\to$ 77 .3	0 .67 $\to$ 0 .71	1 .4 $\to$ 1 .7	5 .3 $\to$ 2 .0	4 .0	0 .0 $\to$ 0 .0	0 .0 $\to$ 0 .0	4 .7 $\to$ 2 .7

Table 2: Physical indicators under camera faults. Paired entries report results in the order of fault introduction: episode start

\rightarrow

after gripper closure, with corresponding clean references. NTC reports the percentage of steps; SR, OD, Drop,

F_{5}

, JVE, and JLP report percentages of episodes. Drop is reported only for faults introduced after gripper closure.

\|\Delta xyz\|

denotes mean translation-command magnitude in native action units. Metric definitions and measurement protocols appear in Appendix B .

Figure 3: Real-world outcomes on the WidowX block-in-bowl task, with ten trials per condition and faults introduced from episode start. Left: Task success; labels give successful trial counts and error bars show Wilson 95% confidence intervals. Right: Operator-annotated non-target contact and drop incidence, and the median minimum joint-limit margin across trials. Unlike simulation NTC , contact is measured per trial. Joint-limit margin measures the minimum distance to configured arm-joint limits in radians; smaller values indicate closer proximity.

Panel of Figure 4 — Figure 4: Task and physical outcomes of selective content removal on Spatial (150 episodes per cell). (a,b) Task success after removal at episode start. (c) Top: NTC under wrist-object removal versus rendering control. Bottom: JLP after removing scene arm links alone or together with the palm. Both physical indicators use episode-start interventions and up to 105 executed steps. Yes/No and $+$ P/ $-$ P indicate policies with/without proprioception. Exact success values appear in Table 10 ; measurement protocols appear in Appendix B .

Panel of Figure 5 — Figure 5: Effects of blackout training on $\pi 0.5$ on Spatial, with 150 episodes per condition per run. (a) Whiskers show observed ranges across two training runs where available. Full results appear in Table 14 . (b) Physical indicators use the first 105 executed steps; NTC and JVE are step percentages, and OD is an episode percentage.

Panel of Figure 6 — Figure 6: Mean-embedding replacement under episode-start camera faults on Spatial (150 episodes per arm). (a) Task success for $\pi 0.5$ and GR00T. (b) Physical outcomes for $\pi 0.5$ .

Appendix figures & tables31 assets

Supplementary material from the paper’s appendix.

Appendix

Setting	$\pi 0.5$	GR00T
Initialization	Base $\pi 0.5$ weights	GR00T-N1.7-3B base weights
Evaluated update	10,000	4,000
Effective batch size	32	640
Optimizer	AdamW	AdamW
Learning rate	Linear warmup to $5\times 10^{-5}$	$10^{-4}$ peak, cosine schedule
Warmup	10,000 updates	200 updates (5% of training)

Appendix

Table 3: Training settings for the principal simulation policies. The with- and without-proprioception policies are trained separately within each model family. These settings do not describe the released or short-budget checkpoints used in the additional mitigation experiments.

Setting	Value
Control frequency	20 Hz
Initial settling	Ten simulator steps before policy execution
Replanning interval	Five executed actions for $\pi 0.5$ ; eight for GR00T
Denoising steps	Ten for $\pi 0.5$ ; four for GR00T
Episode limit	Spatial: 220; Object: 280; Goal: 300 executed steps
Default evaluation size	Ten tasks $\times$ 15 initial states = 150 episodes per suite and condition

Appendix

Table 4: Closed-loop simulation evaluation settings. Episode limits apply before any separately requested post-completion recording. Physical measurements use the windows specified in Appendix B .

Model	Proprio.	Clean	Scene black	Wrist black	Scene frozen	Wrist frozen
$\pi 0.5$	Yes	93.1	28.9	13.1	5.8	3.6
$\pi 0.5$	No	94.4	9.1	1.8	0.2	0.0
GR00T	Yes	98.7	51.6	12.2	9.3	0.0
GR00T	No	97.1	34.4	0.0	5.1	0.0

Appendix

Table 5: Task success (%) for policies with and without proprioception. Both models average Spatial, Object, and Goal (

n=450

per condition).

Model	Proprio.	Condition	Drop ( $n$ )	Grasp ( $n$ )	Drop (%)	Drop / grasp (%)
Camera faults
$\pi 0.5$	Yes	Clean	2	146	1.3	1.4
$\pi 0.5$	Yes	Scene blackout	3	137	2.0	2.2
$\pi 0.5$	Yes	Scene freezing	8	144	5.3	5.6
$\pi 0.5$	Yes	Wrist blackout	6	135	4.0	4.4
$\pi 0.5$	Yes	Wrist freezing	3	138	2.0	2.2

Appendix

Table 6: Drop counts and grasp opportunity under camera faults. Every row contains 150 episodes. The reported Drop rate uses all 150 episodes; the conditional rate uses only episodes with an established grasp. A grasp requires bilateral finger contact and target height more than 2 cm above its first recorded height for three consecutive steps. Camera faults are introduced after gripper closure, with matched clean references.

Model	Proprio.	Condition	Grasp	Prefix	Full post50	Zero post50
$\pi 0.5$	Yes	Wrist blackout	135	132.0	117	0
$\pi 0.5$	Yes	Wrist freezing	138	135.0	126	0
$\pi 0.5$	No	Wrist blackout	130	220.0	145	1
$\pi 0.5$	No	Wrist freezing	134	220.0	136	1
GR00T	Yes	Wrist blackout	142	183.5	125	0
GR00T	Yes	Wrist freezing	144	220.0	141	0

Appendix

Table 7: Measurement opportunity in the representative contrasts. Every row has 150 episodes. Prefix is the median number of recorded steps through first completion (full trace for failures). Full post50 counts episodes with all 50 observed steps after logged fault onset before completion; zero post50 counts episodes with no such exposure. Episodes with shorter exposure remain in the all-episode denominator. Grasp is the number satisfying the established-grasp rule before completion.

Model	Proprio.	Metric	A	B	Primary $\Delta$ [95% CI]	$\Delta_{105}$	$\Delta_{\mathrm{post50}}$	$\Delta_{\mathrm{full}}$
A: wrist blackout; B: wrist freezing (after gripper closure)
$\pi 0.5$	Yes	Drop	4.0	2.0	+2.0 [+0.0, +4.7]	+2.7	+2.7	+2.0
$\pi 0.5$	Yes	JVE	0.0	6.7	-6.7 [-14.0, -0.7]	+0.0	+0.0	-6.7
$\pi 0.5$	No	Drop	28.7	6.7	+22.0 [+8.0, +38.7]	+4.7	+6.7	+22.0
$\pi 0.5$	No	JVE	3.3	30.7	-27.3 [-52.7, -4.7]	-0.7	+0.0	-27.3
GR00T	Yes	Drop	6.7	4.0	+2.7 [-2.0, +7.3]	+2.0	+2.0	+2.7

Appendix

Table 8: Uncertainty and measurement-window sensitivity for representative physical contrasts. Each contrast uses 150 paired Spatial episodes. A and B name the two conditions in each block. Rates are percentages of episodes, and

\Delta=A-B

is an absolute difference on that scale. Primary intervals are 95% task-cluster bootstrap intervals from 10,000 resamples of the ten tasks. The primary window ends at first task completion, or at episode end for failures. Additional columns give differences for the first 105 steps, the first 50 steps from logged fault onset (both capped at first completion), and the full recorded trace. Drop uses the same contact-conditioned rule in every column; both contact loss and the qualifying descent must lie inside the selected window.

Panel of Figure 7 — Figure 7: $\pi 0.5$ executed action coordinates under clean inputs, scene blackout, and wrist blackout. Rows follow the clean first-close order; gray marks time after episode termination. Color scales are shared across conditions and panels for each coordinate.

Panel of Figure 8 — Figure 8: GR00T executed action coordinates under clean inputs, scene blackout, and wrist blackout. Rows follow the clean first-close order; gray marks time after episode termination. Color scales are shared across conditions and panels for each coordinate.

Policy	Proprio.	Scene black	Scene frozen	Wrist black	Wrist frozen
$\pi 0.5$	Yes	$.94\to.27$	$1.00\to.33$	$.99\to.74$	$1.00\to-.10$
$\pi 0.5$	No	$.85\to.15$	$1.00\to.24$	$.97\to.55$	$.99\to.11$
GR00T	Yes	$.99\to.46$	$1.00\to.33$	$.98\to.76$	$.99\to.07$
GR00T	No	$.99\to.31$	$1.00\to.12$	$.50\to.06$	$.99\to.13$

Appendix

Table 9: Median xyz command cosine relative to paired clean execution. Each cell shows early

[0,20)

\rightarrow

later

[50,80)

values over eligible episode-step pairs at

\epsilon=0.05

. A value near 1 indicates similar native command direction, irrespective of magnitude.

Figure 9: Immediate

\pi 0.5

gripper commands for 200 paired Spatial observations per phase, comparing clean inputs with both camera images blacked out. Commands are averaged over the planned five-step execution prefix. Open is on the left and close is on the right.

Scene view
	$\pi 0.5$		GR00T
Proprioception	Yes	No	Yes	No
Robot rendering control	95.3 (143)	94.0 (141)	97.3 (146)	94.0 (141)
Fingers hidden	95.3 (143)	96.0 (144)	96.0 (144)	96.7 (145)
Arm links hidden	93.3 (140)	90.7 (136)	96.7 (145)	88.7 (133)
Arm links and fingers hidden	89.3 (134)	75.3 (113)	92.7 (139)	81.3 (122)

Appendix

Table 10: Exact task success for the selective-removal conditions in Figure 4 . Cells show success percentage (successful episodes), with 150 LIBERO-Spatial episodes per condition. Robot rendering controls remove shadows; object rendering controls retain them. The other camera is unmodified.

Removed content	Policy	Proprio.	Success	$\Delta xyz$	NTC	OD
Scene robot	$\pi 0.5$	Yes	.507	.034	.055	.133
	$\pi 0.5$	No	.133	.084	.082	.247
	GR00T	Yes	.467	.049	.029	.073
	GR00T	No	.233	.075	.045	.200
Wrist objects	$\pi 0.5$	Yes	.027	.061	.095	.200
	$\pi 0.5$	No	.027	.110	.126	.327

Appendix

Table 11: Task, command, and physical outcomes after selective content removal. Translation reports RMS change from the matched rendering control over the first 20 executed steps. Physical indicators use up to 105 steps: NTC is non-target contact, and OD is sustained non-target object displacement. Each row contains 150 episodes.

Panel of Figure 10 — Figure 10: Camera-disagreement inputs and policy responses. (a) Scene/wrist input pairs with matching views (outer columns) or different times or target object positions (middle columns). (b) Scene- and wrist-reference following across execution stages, averaged over both disagreement conditions. (c) Median translation-action projections toward the shifted-target reference, grouped by end-effector distance.

Model	Proprio.	Phase	$n$	Scene (%)	Wrist (%)
$\pi 0.5$	Yes	Early episode	61	23.0	77.0
$\pi 0.5$	Yes	Near close	100	41.5	58.5
$\pi 0.5$	Yes	Transport	100	61.5	38.5
$\pi 0.5$	Yes	Near episode end	65	27.7	72.3
$\pi 0.5$	No	Early episode	61	30.3	69.7
$\pi 0.5$	No	Near close	100	34.5	65.5

Appendix

Table 12: Camera-reference following at a 10-frame lag, pooled equally over the B/A and A/B input pairs in Figure 10(b) . Each of

n

selected observations contributes two queries. Shares use the closer full-seven-action reference; no exact ties occur. Within each policy and phase, selection retains reference separation at or above the median.

Model	Proprio.	Distance (cm)	$n$	Scene median	Wrist median
$\pi 0.5$	Yes	$[0,6)$	1119	0.030	0.874
$\pi 0.5$	Yes	$[6,12)$	569	0.115	0.772
$\pi 0.5$	Yes	$[12,24)$	386	0.268	0.706
$\pi 0.5$	Yes	$[24,40)$	166	0.377	0.613
$\pi 0.5$	No	$[0,6)$	1125	0.020	0.948
$\pi 0.5$	No	$[6,12)$	627	0.071	0.881

Appendix

Table 13: Median XYZ action projection toward the displaced-target reference by end-effector distance. Distance bins are in cm and use

[\mathrm{lo},\mathrm{hi})

;

n

is selected probes per bin and is shared by scene and wrist within a row.

Recipe	Proprio.	$n_{\mathrm{runs}}$	Clean	Scene black	Scene frozen
None	Yes	1	93.3	22.0	2.7
None	No	1	93.3	14.0	1.3
Scene 10%	Yes	2	96.0 [95.3,96.7]	91.0 [89.3,92.7]	34.3 [30.0,38.7]
Scene 10%	No	1	92.7	86.0	14.7
Scene 20%	Yes	2	94.7 [94.0,95.3]	94.0 [93.3,94.7]	53.0 [45.3,60.7]
Scene 20%	No	1	95.3	91.3	38.0

Appendix

Table 14: Task success (%) for the evaluated

\pi 0.5

LIBERO-Spatial training recipe and proprioception condition. Entries are mean [minimum, maximum] across training runs; the dedicated

n_{\mathrm{runs}}

column gives the number of runs.

Suite	Proprio.	Training	Clean	Scene black	Scene frozen	Wrist black
Spatial	Yes	Plain	99.3	44.7	7.3	8.0
Spatial	Yes	Scene blackout	98.7	99.3	58.0	7.3
Spatial	Yes	Wrist blackout	98.7	34.7	0.0	93.3
Spatial	Yes	Scene + wrist	100.0	96.0	7.3	88.0
Spatial	No	Plain	97.3	32.7	0.7	0.0
Spatial	No	Scene blackout	99.3	97.3	6.7	0.0

Appendix

Table 15: GR00T task success at episode-start faults across Spatial, Goal, and Object (%; 150 rollouts per cell). Yes and No indicate policies with and without proprioception. Training uses 20% camera blackout. One training seed per recipe.

Transplanted component	Success (%)
None	22.7
Vision encoder + projector	38.0
VLM backbone	90.0
Action expert	26.0
All components (trained policy)	94.7

Appendix

Table 16: Component transplants from a scene-blackout-trained

\pi 0.5

policy with proprioception into the corresponding policy trained without blackout.

Proprio.	Camera	Fault	Policy/input	SR	NTC	JVE	OD
Yes	scene	black	Control / clean	0.953	0.025	0.000	0.013
Yes	scene	black	Control / fault	0.233	0.054	0.000	0.053
Yes	scene	black	Trained / fault	0.953	0.021	0.000	0.027
Yes	scene	frozen	Control / clean	0.953	0.025	0.000	0.013
Yes	scene	frozen	Control / fault	0.027	0.101	0.018	0.253
Yes	scene	frozen	Trained / fault	0.587	0.039	0.000	0.060

Appendix

Table 17: Task success and physical outcomes after blackout training on LIBERO-Spatial. SR is whole-episode success; contact and joint-speed exceedance are step rates, and displacement reports sustained movement of a non-target object. Physical measurements use up to 105 steps.

Figure 11:

\pi 0.5

task success and physical indicators after blackout training on Object. Open circles, squares, and dashed lines denote base-fault, camera-trained-fault, and base-clean results. Physical indicators use the first 105 executed steps. NTC and JVE report step percentages; OD reports episode percentages.

Figure 12:

\pi 0.5

task success and physical indicators after blackout training on Goal. Open circles, squares, and dashed lines denote base-fault, camera-trained-fault, and base-clean results. Physical indicators use the first 105 executed steps. NTC and JVE report step percentages; OD reports episode percentages. NTC and OD use the 120 episodes with a defined target out of 150 rollouts per condition.

Figure 13:

\pi 0.5

task success with proprioception under detector-triggered mean-embedding replacement. Open circles denote the unmodified policy and squares the replacement intervention. Each comparison uses 150 Spatial episodes paired within one evaluation setup.

Fault	Original	Replacement	$\Delta$ success (%) [95%]	McNemar $p$
Scene start · Black	34/150	97/150	+42.0 [+26.7, +58.7]	3.4e-15
Scene start · Frozen	9/150	102/150	+62.0 [+50.7, +73.3]	2.0e-28
Wrist start · Black	35/150	36/150	+0.7 [-10.0, +11.3]	1.000
Wrist start · Frozen	0/150	39/150	+26.0 [+14.7, +38.0]	3.6e-12
Scene Random · Black	67/150	111/150	+29.3 [+5.3, +52.7]	1.7e-07
Scene Random · Frozen	15/150	108/150	+62.0 [+48.0, +74.0]	6.0e-26

Appendix

Table 18: Task success without and with detector-triggered mean-embedding replacement for camera faults from episode start and random, grasp, and hold onsets. Start denotes episode start. Each paired LIBERO-Spatial condition has 150 episodes. Differences are absolute changes in success (%) with 95% task-cluster bootstrap intervals;

p

is exact McNemar.

Fault	$\Delta$ NTC (%) [95% CI]	$\Delta$ OD (%) [95% CI]
Scene start · Black	-0.5 [-2.6, +1.8]	+6.7 [+0.7, +13.3]
Scene start · Frozen	-5.0 [-7.5, -2.5]	-15.3 [-23.3, -7.3]
Wrist start · Black	+2.1 [+0.9, +3.5]	+2.0 [+0.0, +4.7]
Wrist start · Frozen	-8.6 [-11.3, -6.1]	-16.0 [-22.0, -10.0]
Scene Random · Black	+0.1 [-0.9, +1.2]	+2.0 [-2.0, +6.7]
Scene Random · Frozen	-4.9 [-7.1, -2.7]	-17.3 [-24.0, -10.7]

Appendix

Table 19: Changes in physical outcomes after detector-triggered mean-embedding replacement for the same paired conditions as Table 18 . Each cell reports replacement minus original with an episode-level 95% bootstrap interval, using up to 105 executed steps. Differences are absolute changes in NTC (% of steps) and OD (% of episodes). OD records non-target displacement above 2 cm for six consecutive samples.

Figure 14: Absolute input-group attention masses for scene vision, wrist vision, and state across clean and camera-fault probes. Values are attention masses multiplied by 100, measured over 96 fixed-probe episodes across four suites. This probe cohort is separate from the LIBERO-Spatial gain rollouts.

Input	State bias	Scene bias	Success (%)
Scene blackout	0	0	24.7
	0.25	0	23.3
	0.5	0	21.3
	1	0	14.7
	2	0	4.0
	4	0	0.0

Appendix

Table 20: Task success under manually selected attention biases. Each row uses 150 Spatial episodes and the same

\pi 0.5

checkpoint with proprioception. Unlisted token groups have zero bias.

Policy or intervention	Scene blackout (%)	Wrist blackout (%)
Unmodified control	22.0	21.3
Full group adjustment	26.7	22.7
Camera-only adjustment	—	26.7
Blackout-trained reference	94.7	72.7

Appendix

Table 21: Attention biases calibrated to blackout-trained policies. Success uses 150 episodes per cell. Full adjustment targets state and camera groups; camera-only adjustment targets camera groups. The trained reference is a separate policy trained with blackout in the evaluated camera. A dash denotes a setting not shown in this summary.

Figure 15: Inference-time attention reweighting under matched scene- and wrist-blackout conditions. Each arm contains 150 Spatial episodes; intervals are 95% Wilson confidence intervals. Trained-checkpoint controls remain distinct from inference-only interventions.

Policy	Camera fault	No mask	Expert only	Prefix + expert
Released	Scene freezing	1.3	4.0	47.3
	Wrist freezing	0.0	0.0	—
Fine-tuned, with proprio.	Scene freezing	25.3	30.0	18.0
	Wrist freezing	0.0	0.0	—
Fine-tuned, without proprio.	Scene freezing	6.7	14.0	10.7
	Wrist freezing	0.0	0.0	—

Appendix

Table 22: Task success (%) under oracle camera-token masking, with 150 episodes per cell. The two fine-tuned policies are a separate non-EMA checkpoint pair. All faults and masks start at episode start. Dashes indicate unevaluated settings.

Evaluation input	Control	Augmentation	Augmentation + freshness
Clean	42.0	36.0	46.0
Scene freezing	4.0	22.0	12.0
Wrist freezing	0.0	2.0	10.0

Appendix

Table 23: Task success (%) after 4,000 updates from the base checkpoint, with 50 episodes per cell. Freezing begins after three consecutive gripper-close commands. All scheduled episodes remain in the denominator, including those that do not reach the closing trigger. The control receives no temporal augmentation.

Evaluation input	Control	Delay 8–40	Delay 2–8	+ Consistency
		Temporal mixture
Clean	98.7	98.7	98.7	73.3
Scene freezing	0.0	7.3	1.3	4.7
Wrist freezing	0.0	0.0	0.0	0.0
Wrist blackout	0.0	0.0	0.0	0.0

Appendix

Table 24: Task success (%) after 4,000 additional updates from the released LIBERO policy, with 150 episodes per cell. Both temporal mixtures contain landmark freezing and delayed images; only the sampled delay range differs. Consistency uses the 8–40-frame mixture. Faults begin at episode start.

Explore similar work

Uncovering Vulnerability of Vision-Language-Action Models under Joint-Level Physical Faults
Jun 9, 2026Minsoo Jo, Taeju Kwon, Junha Chun +2Diffusion-Based Vision-Language-Actions Robot Failures
Graduate School of Data Science, Seoul National University
ProbeAct: Probe-Guided Training-Free Failure Recovery in Vision-Language-Action Models
Jun 8, 2026Fan Zhang, Seongbin Park, Baharan Mirzasoleiman +2Automata
University of California Los Angeles United States
ActFovea: Runtime Safeguarding for VLA Policies via Spatiotemporal Visual-Action Consistency
Jul 31, 2026Wenda Yu, Tianshi Wang, Fengling Li +3Robotic Manipulation Runtime Safety Filtering
Tongji University · Mohamed bin Zayed University of Artificial Intelligence · Shanghai Artificial Intelligence Laboratory +1

Jun 9, 2026, cs.ROJ/K move · Enter open · S save

Uncovering Vulnerability of Vision-Language-Action Models under Joint-Level Physical Faults

Minsoo Jo, Taeju Kwon, Junha Chun, Youngjoon Jeong+1

Graduate School of Data Science, Seoul National University

Diffusion-Based Vision-Language-Actions Robot Failures Robotic Systems Embodiment Correction

Open paper AskPDF arXiv

Deploying Vision-Language-Action (VLA) models in real robotic systems requires robustness not only to semantic and perceptual variations, but also to embodiment-side faults that change how actions are physically realized. Real robots can experience joint-level changes caused by actuator degradation, hardware faults, safety limits, collision damage, or wear-induced friction. These faults are critical because they alter the action-to-motion interface of a policy, disrupting the learned closed-loop relationship between commanded actions, realized motion, and subsequent observations. In this work, we study realistic joint-level physical faults and show that VLA models are vulnerable when predicted actions are executed through a perturbed robot body. Our analysis reveals joint-dependent effects, with heterogeneous degradation in task success across affected joints. We also show that performance drops cannot be attributed solely to physical infeasibility, since feasible faults such as increased joint friction can still substantially reduce success rates and induce closed-loop execution mismatch. Motivated by these findings, we propose Joint-level Physical-fault Aware Residual Calibrator (J-PARC), a lightweight residual calibration framework built on top of a frozen VLA policy. J-PARC infers a latent joint-fault regime from recent joint dynamics and conditions a shared residual calibrator on this regime, enabling adaptive action correction across faulty joints. Experiments show that J-PARC improves robustness under joint-level faults while preserving fault-free environment performance.

Abstract

Explore similar work

Uncovering Vulnerability of Vision-Language-Action Models under Joint-Level Physical Faults

ProbeAct: Probe-Guided Training-Free Failure Recovery in Vision-Language-Action Models

ActFovea: Runtime Safeguarding for VLA Policies via Spatiotemporal Visual-Action Consistency