HalluPrism: When Multimodal Uncertainty Should Diagnose, Not Decide

Aman Prakash; Sourish Dasgupta; Tanmoy Chakraborty

cs.LGAug 29, 2026

HalluPrism: When Multimodal Uncertainty Should Diagnose, Not Decide

Authors: Aman Prakash, Sourish Dasgupta, Tanmoy Chakraborty

Organizations: LCS2 Lab, Indian Institute of Technology Delhi, New Delhi, India · KDM Lab, Dhirubhai Ambani University, Gujarat, India

Abstract

Multimodal Large Language Models (MLLMs) can assign similar confidence to answers that fail for different reasons. We propose HalluPrism, a behavioral diagnostic that re-runs an answer after visual degradation, blank-image replacement, and grounding or relation checks. These targeted probes yield a signature over visual-perturbation sensitivity (V ), image-removal confidence retention (L), and grounding/relation-probe instability (A). Across 58K+ examples from four benchmarks and four MLLMs, image-removal confidence retention is most prevalent, while grounding/relation-probe instability better separates failure families. Only 18 of 48 source-target checks are diagonally aligned, so the coordinates should be interpreted jointly rather than as independent causal sources. With the dataset fixed, the joint signature improves failure-family AUROC from 0.634 to 0.769 on HallusionBench and from 0.707 to 0.817 on VizWiz, with smaller gains on POPE and VSR. In pooled XGBoost analysis, AUROC rises from 0.78 with scalar confidence to 0.95 with (V, L, A) and 0.97 when confidence is added. The same signature does not automatically improve correctness ranking. The three tested direct scalarizations can harm it. These results separate failure diagnosis from abstention scoring: multimodal uncertainty should characterize failure structure before it is used to decide whether to abstain or correct.

Figures & tables

Figure 1: An overview of the HalluPrism perturbation-response protocol . Here $L$ denotes image-removal confidence retention; language-prior reliance is only a possible interpretation of this response. The scores are behavioral probes, not a causal decomposition. Full prompt templates and probe parameters are given in Appendix 9 .

Dataset	N Per	Primary Stress	Source Probe Role
HallusionBench	951	Visual illusion, attribute error	Visual and alignment pressure
POPE	9,000	Object-presence hallucination	Blank-image and object-prior pressure
VSR	340	Spatial and relation binding	Relation-swap alignment pressure
VizWiz-VQA	4,319	Visual insufficiency, answerability	Visual degradation and blank-image pressure
Models	Gemma 4, LLaVA-Mistral, LLaVA-Vicuna, and Qwen3-VL under deterministic decoding.

Table 1: Evaluation suite and roles. Evaluation roles focus on mixed failure-family diagnosis for HallusionBench, object hallucination/abstention for POPE, relation error/routing breakdown for VSR, and answerability diagnosis for VizWiz-VQA. Each model-dataset output receives scalar confidence

c

and Perturbation-Response Signature

(V,L,A)

. Full dataset roles, split details, probe availability, and failure-family mappings are given in Appendix 8 .

Evidence Block	Setting	Main Value
Mean image-removal confidence retention	All model-dataset settings	$L$ ranges from $0.91$ to $1$
Blank-image match	All model-dataset settings	LP match ranges from $37.6\%$ to $96.5\%$
Highest visual-perturbation sensitivity	VizWiz settings	$V$ ranges from $0.19$ to $0.30$
Highest alignment instability	Gemma 4/VSR, Qwen3-VL/VSR	$A=0.68$ and $A=0.61$
Raw dominant source	$58{,}440$ pooled samples	$L/A/V=58{,}287/151/2$
Magnitude-based diagonal pass count	All intervention checks	$18/48$ pass

Table 2: Source-profile and Entanglement Summary . LP match is exact-answer agreement under blank-image replacement. Diagonal pass counts test whether the intended source has the largest absolute response under the corresponding ruined condition. Full profiles and intervention deltas are given in Appendix Tables 13 - 19 .

Failure-Family Diagnosis
Classifier	Scalar	$\mathbf{V+L+A}$	$\mathbf{V+L+A+c}$
Logistic Regression	$0.75$	$0.83$	$\mathbf{0.87}$
XGBoost	$0.78$	$0.95$	$\mathbf{0.97}$
LightGBM	$0.78$	$0.95$	$\mathbf{0.97}$

Table 3: Failure-family diagnosis, feature importance, and dataset-ID controls. Classifier rows report pooled weighted one-vs-rest AUROC over

58{,}440

examples. Feature rows report pooled tree-based feature importance. Dataset-ID controls use LightGBM and report gains after adding source features beyond dataset identity

D

, model identity

M

, and scalar confidence

c

. Full-precision diagnosis and control results are given in Appendix Tables 21 - 25 .

Control	Setting	AUROC	Macro-F1
$L_{\mathrm{conf}}$	clipped ratio	$0.828$	$0.333$
$L_{\mathrm{match}}$	match only	$0.823$	$0.346$
$L_{\mathrm{stable}}$	floored denominator	$0.823$	$0.349$
$c$	excluding VSR	$0.761$	$0.329$
$(V,L,A)$	excluding VSR	$0.839$	$0.422$

Table 4: Probe-design controls. The image-removal rows replace only the

L

coordinate inside the pooled

(V,L,A)

diagnosis feature set. The no-VSR control removes the only relation-swap-eligible dataset and compares scalar confidence with

(V,L,A)

under the same non-VSR scope. Details are in Appendix Tables 10 - 11

Abstention: Scalar Confidence Vs. Best Source-Aware Scalarization
Model Family	Metric	Scalar Confidence	Best Source-Aware
Gemma 4	AUROC	$0.81$	$0.77\ (0.04\downarrow)$
LLaVA-Mistral	AUROC	$0.84$	$0.81\ (0.03\downarrow)$
LLaVA-Vicuna	AUROC	$0.83$	$0.78\ (0.05\downarrow)$
Qwen3-VL	AUROC	$0.81$	$0.73\ (0.08\downarrow)$

Table 5: Decision-facing behavior . Abstention compares scalar confidence against the best source-aware scalarization for each model family. Routing compares no intervention, generic prompting, and matched routing. The Highest accuracy row counts which policy obtains the best accuracy in each model-dataset setting. The pairwise rows compare generic prompting and matched routing across the same

16

settings. Full-precision results appear in Appendix Tables 32 - 36 .

Dataset	Split / Subset	N Per Model	Primary Diagnostic Stress	Role In HalluPrism
HallusionBench	Image-paired subset	$951$	Visual illusion, attribute error, and grounded hallucination	Mixed failure-family diagnosis and routing analysis under visually grounded hallucination cases.
POPE	Random, popular, and adversarial splits	$9{,}000$	Object-presence hallucination and object priors	Object-hallucination diagnosis, question-only prior pressure, and abstention boundary check.
VSR	Zero-shot development examples	$340$	Spatial and relation binding	Relation-swap alignment probing and routing break-rate analysis.
VizWiz-VQA	Validation split	$4{,}319$	Visual insufficiency and answerability	Answerability diagnosis, visual-fragility analysis, and abstention boundary check.

Table 6: Dataset details for the HalluPrism evaluation suite. Counts are per model before pooling across the four evaluated model families.

Dataset	Output Space	Incorrect-Prediction Label	Diagnostic Interpretation
HallusionBench	yes/no	Attribute error or object hallucination	Mixed visually grounded hallucination and illusion failure.
POPE	yes/no	Object hallucination	Object-presence error under random, popular, or adversarial object-prior pressure.
VSR	true/false	Spatial or relation error	Cross-modal binding failure over explicit spatial relations.
VizWiz-VQA	Short answer or unanswerable	Answerability failure	Visual insufficiency or failure to respect answerability constraints.

Table 7: Failure-family mapping used for diagnostic classification. Correct predictions are mapped to no failure . Incorrect predictions are mapped using each benchmark’s primary stress factor and available labels.

Dataset	Per-Model N	Pooled N	Failure-Family Counts Per Model
HallusionBench	951	3,804	attribute = 360, object hallucination = 591
POPE	9,000	36,000	object hallucination = 9,000
VizWiz-VQA	4,319	17,276	answerability = 1,385, none = 2,934
VSR	340	1,360	spatial = 340

Table 8: Dataset composition and failure-family coverage. Counts are reported per model and after pooling across the four evaluated model families.

Dataset	Visual Degradation	Blank Image	Grounding Re-Check	Relation Swap
HallusionBench	Yes	Yes	Yes	No
POPE	Yes	Yes	Yes	No
VSR	Yes	Yes	Yes	Yes
VizWiz-VQA	Yes	Yes	Yes	No

Table 9: Probe availability across datasets. Relation-swapped probing is used only when the input contains an explicit perturbable relation.

Variant	Removal Input	Ratio Treatment	$L$ Dom.	AUROC	Macro-F1
$L_{\mathrm{conf}}$	Blank	Clipped ratio	$99.7\%$	$0.828$	$0.333$
$L_{\mathrm{persist}}$	Blank	Match-gated ratio	$51.6\%$	$0.823$	$0.349$
$L_{\mathrm{match}}$	Blank	None	$51.7\%$	$0.823$	$0.346$
$L_{\mathrm{stable}}$	Blank	Floored denominator	$51.6\%$	$0.823$	$0.349$

Table 10: Image-removal saturation controls.

L

Dom. denotes the fraction of examples for which

L

is the raw dominant coordinate. AUROC and Macro-F1 use the same pooled failure-family diagnosis setting with a stratified 5-fold logistic-regression classifier, replacing only the image-removal coordinate.

Feature Set	Evaluation Scope	Relation-Swap Eligible	AUROC	Macro-F1
$c$	All datasets	Yes	$0.752$	$0.260$
$(V,L,A)$	All datasets	Yes	$0.828$	$0.333$
$c$	Excluding VSR	No	$0.761$	$0.329$
$(V,L,A)$	Excluding VSR	No	$0.839$	$0.422$

Table 11: Alignment availability control. We compare scalar confidence

c

with the Perturbation-Response Signature

(V,L,A)

under the same stratified 5-fold logistic-regression failure-family diagnosis setting. The no-VSR setting removes the only relation-swap-eligible dataset, so any remaining gain cannot be attributed to the VSR-only relation branch or to the grounding-relation weight in Eq. 6 .

Setting	Blur	Crop	Bright.	Noise	$\mathbf{\bar{V}}$	Flip
Weak	$8$	$75\%$	$50\%$	$10$	$0.132$	$0.174$
Default	$15$	$50\%$	$30\%$	$25$	$0.196$	$0.263$
Strong	$22$	$25\%$	$15\%$	$50$	$0.252$	$0.342$

Table 12: Visual-perturbation severity analysis. Stronger perturbations increase both mean visual sensitivity and answer-flip rate.

Model	Dataset	$N$	Acc	$\bar{V}$	$\bar{L}$	$\bar{A}$	LP Match	Dom- $V$	Dom- $L$	Dom- $A$
Gemma 4	HallusionBench	951	68.2%	0.122	0.958	0.501	56.9%	0	949	2
Gemma 4	POPE	9000	86.9%	0.114	0.999	0.336	53.3%	0	9000	0
Gemma 4	VizWiz	4319	54.0%	0.303	0.996	0.638	37.5%	2	4308	9
Gemma 4	VSR	340	71.8%	0.143	0.999	0.675	64.4%	0	340	0
LLaVA-Mistral	HallusionBench	951	52.6%	0.090	0.979	0.241	58.4%	0	951	0
LLaVA-Mistral	POPE	9000	89.1%	0.104	0.999	0.061	55.8%	0	9000	0

Table 13: Full source-profile with accuracy and raw dominant-source ( Dom ) counts. LP Match denotes exact-answer agreement between the original image and blank-image condition.

Model	$N$	Dominant $V$	Dominant $L$	Dominant $A$
Gemma 4	$14{,}610$	$2$ ( $0.01\%$ )	$14{,}597$ ( $99.91\%$ )	$11$ ( $0.08\%$ )
LLaVA-Mistral	$14{,}610$	$0$ ( $0.00\%$ )	$14{,}593$ ( $99.88\%$ )	$17$ ( $0.12\%$ )
LLaVA-Vicuna	$14{,}610$	$0$ ( $0.00\%$ )	$14{,}489$ ( $99.17\%$ )	$121$ ( $0.83\%$ )
Qwen3-VL	$14{,}610$	$0$ ( $0.00\%$ )	$14{,}608$ ( $99.99\%$ )	$2$ ( $0.01\%$ )
Pooled	$58{,}440$	$2$ ( $<0.01\%$ )	$58{,}287$ ( $99.74\%$ )	$151$ ( $0.26\%$ )

Table 14: Largest-Coordinate Counts . We compute the largest coordinate using

\arg\max(V,L,A)

. The largest coordinate is overwhelmingly

L

because

L

is near saturation in most settings; this count should therefore be read as prevalence, not diagnostic value.

Model	Visual-Ruined	Language-Ruined	Alignment-Ruined	All Checks
Gemma 4	3/4	1/4	1/4	5/12
LLaVA-Mistral	3/4	0/4	1/4	4/12
LLaVA-Vicuna	1/4	0/4	1/4	2/12
Qwen3-VL	4/4	1/4	2/4	7/12
Pooled	11/16	2/16	5/16	18/48

Table 15: Magnitude-based diagonal-dominance pass counts for the interventional entanglement test . A check passes when the intended source coordinate has the largest absolute change under the corresponding ruined condition.

Dataset	Intervention	$\Delta V$	$\Delta L$	$\Delta A$	Outcome
HallusionBench	Visual-ruined	-0.075	+0.016	+0.126	FAIL
HallusionBench	Language-ruined	-0.061	-0.025	+0.088	FAIL
HallusionBench	Alignment-ruined	-0.001	-0.000	-0.001	PASS
POPE	Visual-ruined	-0.086	+0.000	+0.068	PASS
POPE	Language-ruined	-0.031	-0.052	+0.247	FAIL
POPE	Alignment-ruined	+0.000	+0.000	+0.000	FAIL

Table 16: Interventional Entanglement

\Delta

for Gemma 4 . A check passes when the intended source coordinate has the largest absolute change under the corresponding ruined condition. Resultsare computed using unrounded

\Delta

Dataset	Intervention	$\Delta V$	$\Delta L$	$\Delta A$	Outcome
HallusionBench	Visual-ruined	-0.026	+0.011	+0.012	PASS
HallusionBench	Language-ruined	-0.051	+0.007	-0.054	FAIL
HallusionBench	Alignment-ruined	+0.001	+0.000	+0.003	PASS
POPE	Visual-ruined	-0.050	+0.001	-0.006	PASS
POPE	Language-ruined	-0.076	-0.014	+0.097	FAIL
POPE	Alignment-ruined	+0.000	+0.000	+0.000	FAIL

Table 17: Interventional Entanglement

\Delta

for LLaVA-Mistral . A check passes when the intended source coordinate has the largest absolute change under the corresponding ruined condition. Results are computed using unrounded

\Delta

Dataset	Intervention	$\Delta V$	$\Delta L$	$\Delta A$	Outcome
HallusionBench	Visual-ruined	-0.020	+0.038	+0.022	FAIL
HallusionBench	Language-ruined	-0.053	+0.037	-0.024	FAIL
HallusionBench	Alignment-ruined	+0.002	-0.001	-0.002	PASS
POPE	Visual-ruined	-0.029	+0.008	+0.239	FAIL
POPE	Language-ruined	-0.101	-0.004	-0.018	FAIL
POPE	Alignment-ruined	+0.000	+0.000	+0.000	FAIL

Table 18: Interventional Entanglement

\Delta

for LLaVA-Vicuna . A check passes when the intended source coordinate has the largest absolute change under the corresponding ruined condition. Results are computed using unrounded

\Delta

Dataset	Intervention	$\Delta V$	$\Delta L$	$\Delta A$	Outcome
HallusionBench	Visual-ruined	-0.085	+0.011	-0.009	PASS
HallusionBench	Language-ruined	-0.047	-0.086	+0.175	FAIL
HallusionBench	Alignment-ruined	-0.001	+0.000	+0.002	PASS
POPE	Visual-ruined	-0.049	+0.000	+0.024	PASS
POPE	Language-ruined	+0.003	-0.149	+0.293	FAIL
POPE	Alignment-ruined	+0.000	+0.000	+0.000	FAIL

Table 19: Interventional Entanglement

\Delta

for Qwen3-VL . A check passes when the intended source coordinate has the largest absolute change under the corresponding ruined condition. Results are computed using unrounded

\Delta

Failure Family	$N$	$\bar{V}$	$\bar{L}$	$\bar{A}$	Scalar	Acc
Answerability	5540	0.124	0.997	0.479	0.880	77.4%
Attribute	1440	0.099	0.946	0.302	0.922	64.3%
None	11736	0.289	0.988	0.587	0.836	45.9%
Object Hallucination	38364	0.107	0.996	0.223	0.967	86.5%
Spatial	1360	0.120	0.977	0.373	0.919	70.2%

Table 20: Pooled source-signature means by failure family.

Model	Classifier	Scalar	V+L+A	V+L+A+Scalar	$\Delta_{VLA-S}$
Gemma 4	Logistic Reg.	0.754 $\pm$ 0.009	0.921 $\pm$ 0.006	0.932 $\pm$ 0.006	+0.167
Gemma 4	XGBoost	0.834 $\pm$ 0.013	0.965 $\pm$ 0.001	0.974 $\pm$ 0.002	+0.131
Gemma 4	LightGBM	0.836 $\pm$ 0.012	0.965 $\pm$ 0.001	0.973 $\pm$ 0.002	+0.130
LLaVA-Mistral	Logistic Reg.	0.847 $\pm$ 0.003	0.892 $\pm$ 0.003	0.907 $\pm$ 0.003	+0.045
LLaVA-Mistral	XGBoost	0.860 $\pm$ 0.003	0.973 $\pm$ 0.002	0.979 $\pm$ 0.002	+0.113
LLaVA-Mistral	LightGBM	0.859 $\pm$ 0.004	0.973 $\pm$ 0.001	0.980 $\pm$ 0.002	+0.115

Table 21: Failure-family prediction AUROC by model and classifier.

\Delta_{VLA-S}

denotes the gain of

V+L+A

over scalar confidence.

Model	Classifier	$V$	$L$	$A$	Scalar Confidence
Gemma 4	XGBoost	0.105	0.305	0.505	0.085
Gemma 4	LightGBM	0.269	0.153	0.292	0.286
LLaVA-Mistral	XGBoost	0.084	0.102	0.688	0.125
LLaVA-Mistral	LightGBM	0.294	0.099	0.331	0.276
LLaVA-Vicuna	XGBoost	0.107	0.178	0.503	0.212
LLaVA-Vicuna	LightGBM	0.211	0.173	0.317	0.299

Table 22: Feature importance for tree-based failure-family diagnosis.

Feature Set	AUROC	Macro-F1
Dataset ID	0.976	0.558
Dataset ID + Model ID	0.975	0.558
Dataset ID + Model ID + Scalar	0.986	0.705
Dataset ID + Model ID + $V+L+A$	0.991	0.783
Dataset ID + Model ID + $V+L+A+\mathrm{Scalar}$	0.991	0.782

Table 23: Dataset-ID controlled failure-family diagnosis using a stratified held-out split. The target is the benchmark-provided failure-family label used in the pooled diagnosis experiment.

Feature Set	AUROC	Macro-F1
Dataset ID	0.735	0.145
Dataset ID + Model ID	0.743	0.145
Dataset ID + Model ID + Scalar	0.847	0.308
Dataset ID + Model ID + $V+L+A$	0.838	0.308
Dataset ID + Model ID + $V+L+A+\mathrm{Scalar}$	0.853	0.348

Table 24: Dataset-ID controlled outcome-family diagnosis using a stratified held-out split. Correct predictions are mapped to no-failure and incorrect predictions to their observed failure family.

Dataset	Target	N	Feature Set	AUROC	F1
HallusionBench	Attribute vs object	3804	Scalar	0.634	0.621
			$V+L+A$	0.769	0.718
			$V+L+A+\mathrm{Scalar}$	0.801	0.739
VizWiz	Answerability vs answerable	17276	Scalar	0.707	0.693
			$V+L+A$	0.817	0.782
			$V+L+A+\mathrm{Scalar}$	0.825	0.784

Table 25: Within-dataset diagnostic controls. Each row fixes the dataset and includes model identity in all feature sets. Improvements from

V+L+A

therefore cannot be attributed to dataset identity alone.

Held-Out Dataset	Model ID + Scalar	$\mathbf{ModelID+V+L+A}$	$\mathbf{ModelID+V+L+A+Scalar}$
HallusionBench	0.339	0.492	0.371
POPE	0.201	0.375	0.229
VSR	0.290	0.416	0.274
VizWiz	0.300	0.328	0.331

Table 26: Leave-dataset-out binary failure detection. This is a stress test only, because multiclass failure-family support differs across datasets.

Dataset	Slice	Expected Source	N	Acc	$\bar{V}$	$\bar{L}$	$\bar{A}$	LP Match
POPE	adversarial	$L$	12000	86.5%	0.119	0.998	0.219	53.8%
POPE	popular	$L$	12000	88.3%	0.108	0.998	0.209	55.6%
POPE	random	$L$	12000	90.3%	0.094	0.997	0.222	57.5%
VizWiz	unanswerable	$V$ /answerability	5540	77.4%	0.124	0.997	0.479	77.9%
VizWiz	answerable	$V$ /answerability	11736	45.9%	0.289	0.988	0.587	22.4%
HallusionBench	not visual-required	$V/A$	2016	56.3%	0.110	0.961	0.318	61.3%

Table 27: Matched source-stress slices from existing metadata. These slices check whether source signatures align with dataset-specific stress factors under fixed task conditions.

Dataset	$N$	Raw Agreement	Fleiss’s $\kappa$	Model-Human Agreement
HallusionBench	50	84.0%	0.290	45.5%
POPE	50	84.0%	0.201	22.2%
VizWiz	50	62.0%	0.086	37.5%
VSR	50	98.0%	0.221	37.1%
All	200	82.0%	0.236	36.3%

Table 28: Human-audit reporting template. Annotators label the dominant failure source as visual, language-prior, alignment, mixed, unclear, or no-failure. We report agreement (Fleiss’s

\kappa

) before adjudication and model-human agreement after adjudication.

Quantity	Value
Examples	200
Annotators	3
Adjudicated disagreement cases	36
Exact all-annotator agreement	20.5%
At least two annotators agree	82.0%
Full-label Fleiss’ $\kappa$	0.236

Table 29: Human-audit agreement and adjudicated label distribution. The audit uses 200 examples, with 50 examples per dataset.

Human Label	Predicted $V$	Predicted $L$	Predicted $A$
Visual	6	7	3
Language-prior	10	6	4
Alignment	18	26	27

Table 30: Human-audit confusion-matrix template for clear-source examples.

Feature Set	Macro-F1	Balanced Acc.	Weighted AUROC
Scalar confidence	0.146	0.229	0.430
Raw $V,L,A$	0.328	0.330	0.597
Normalized $z_{V},z_{L},z_{A}$	0.269	0.310	0.556
Raw $V,L,A$ + scalar	0.322	0.313	0.587

Table 31: Human audit-side clear-source prediction. The classifier is evaluated on the 107 examples adjudicated as visual, language-prior, or alignment.

Model	Scalar	Penalty	Weighted	Geometric	Best SA
Gemma 4	0.809	0.738	0.773	0.741	0.773
LLaVA-Mistral	0.836	0.718	0. 806	0.719	0.806
LLaVA-Vicuna	0.828	0.665	0.778	0.678	0.778
Qwen3-VL	0.809	0.721	0.732	0.721	0.732
Pooled	0.783	0.638	0.767	0.640	0.767

Table 32: Aggregate AUROC for all scalarization variants.

Model	Dataset	Scalar AUROC	Best SA AUROC	$\Delta_{SA-S}$	Scalar Sel@60	Best SA Sel@60
Gemma 4	HallusionBench	0.690	0.605 (weighted)	-0.085	77.5%	73.3%
Gemma 4	POPE	0.779	0.741 (penalty)	-0.038	94.7%	94.7%
Gemma 4	VizWiz	0.734	0.718 (weighted)	-0.016	67.4%	67.7%
Gemma 4	VSR	0.712	0.654 (penalty)	-0.058	81.4%	81.4%
LLaVA-Mistral	HallusionBench	0.609	0.552 (penalty)	-0.057	58.2%	54.9%
LLaVA-Mistral	POPE	0.835	0.807 (weighted)	-0.028	97.5%	95.9%

Table 33: Per-dataset abstention comparison between scalar confidence and the best source-aware scalarization.

Coverage	Scalar Risk	Penalty Risk	Weighted Risk	Geometric Risk	Best SA Risk
5%	0.018	0.290	0.023	0.225	0.023 (Weighted)
10%	0.033	0.194	0.034	0.171	0.034 (Weighted)
20%	0.050	0.134	0.070	0.128	0.070 (Weighted)
40%	0.081	0.094	0.084	0.094	0.084 (Weighted)
60%	0.099	0.157	0.112	0.157	0.112 (Weighted)
80%	0.157	0.224	0.159	0.224	0.159 (Weighted)

Table 34: Pooled risk-coverage values for scalar confidence and source-aware scalarizations.

Model	Dataset	$N$	No Interv.	Generic	Matched	Fix G	Fix M	Break G	Break M
Gemma 4	HallusionBench	951	68.2%	60.9%	68.9%	9.3%	12.6%	15.1%	4.9%
Gemma 4	POPE	9000	86.9%	86.8%	86.3%	11.9%	10.2%	1.9%	2.3%
Gemma 4	VizWiz	4319	54.0%	51.8%	52.2%	5.9%	7.8%	9.2%	9.9%
Gemma 4	VSR	340	71.8%	72.4%	71.5%	22.9%	8.3%	8.2%	3.7%
LLaVA-Mistral	HallusionBench	951	52.6%	50.9%	53.8%	8.2%	12.0%	10.6%	8.4%
LLaVA-Mistral	POPE	9000	89.1%	88.8%	89.2%	6.8%	8.1%	1.2%	0.8%

Table 35: Full intervention-routing results. Generic applies one prompt to all samples. Matched chooses an intervention from the dominant source signature.

Model	Matched $>$ Generic	Matched $>$ No Interv.	Lower Break Rate
Gemma 4	2/4	1/4	2/4
LLaVA-Mistral	4/4	3/4	4/4
LLaVA-Vicuna	2/4	3/4	2/4
Qwen3-VL	2/4	1/4	2/4
Pooled	10/16	8/16	10/16

Table 36: Model-level routing win counts.

Dataset	Sample Size	Primary Stress Factor
HallusionBench	$50$	Visual illusion, attribute error, and grounded hallucination
POPE	$50$	Object-presence hallucination and object priors
VizWiz-VQA	$50$	Visual insufficiency and answerability
VSR	$50$	Spatial and relation binding
Total	$200$	Mixed multimodal failure sources

Table 37: Human-audit sampling plan. The audit uses normalized dominant source for stratification because raw dominance is heavily affected by image-removal confidence-retention saturation.

Category	Shown To Annotators	Hidden From Annotators
Input	Image and question or caption	Perturbed images and probe outputs
Output	Model answer	Scalar confidence $c$
Reference	Gold answer or answerability label	Source scores $(V,L,A)$
Outcome	Correctness flag	Raw and normalized dominant source
Metadata	Dataset name	Routing decision and intervention result

Table 38: Information shown to annotators and information hidden during the audit.

Label	When To Use	Example Pattern
Visual	The image evidence appears weak, degraded, incomplete, too small, occluded, blurry, dark, or otherwise insufficient for the answer.	The model answers an object question, although the relevant object is too blurred or partially outside the frame.
Language-prior	The answer appears plausible from the question alone, common object priors, or default world knowledge, even when the image evidence is weak or absent.	The model answers “yes” to a common object-presence question where the image does not clearly support the object.
Alignment	The relevant entities are visible, while the answer appears to fail because of relation, attribute, grounding, or image-text binding.	The model detects a cone and a car, although it reverses the left-right relation between them.
Mixed	More than one source is plausibly active, and no single source is clearly dominant.	The image is blurry and the question also invites a common prior; either source could plausibly explain the answer.
Unclear	The example does not provide enough information for a reliable judgment, or the provided gold reference answer is factually incorrect.	The image is ambiguous, or the gold answer states there is a TV when the image clearly only contains a monitor.
No Failure	The model prediction accurately matches the correct gold answer, indicating no failure or hallucination occurred.	The model correctly answers the question based on clear visual evidence without error.

Table 39: Human-audit label rubric. The labels are diagnostic descriptions of the observed example, not causal claims about model internals.

Metric	Meaning
Raw agreement	Fraction of examples with identical independent labels.
Fleiss’s $\kappa$	Chance-corrected agreement over the six-label set.
Clear-source coverage	Fraction of examples adjudicated as Visual , Language-prior , or Alignment .
Model-human agreement	Agreement between $s^{*}_{\mathrm{norm}}(x,M)$ and mapped clear-source human labels.
Mixed rate	Fraction adjudicated as Mixed .
Unclear rate	Fraction adjudicated as Unclear .

Table 40: Human-audit reporting fields. We report mixed and unclear cases separately rather than forcing them into one of the three source coordinates.

Dataset / Branch	Examples Per Model	Models	Passes Per Example	Forward Passes
HallusionBench	951	4	7	26,628
POPE	9,000	4	7	252,000
VizWiz-VQA	4,319	4	7	120,932
VSR	340	4	8	10,880
Core Perturbation-Response measurement	14,610	4	–	410,440
Generic and matched routing branches	14,610	4	2	116,880

Table 41: Corrected inference budget for HalluPrism. HallusionBench, POPE, and VizWiz-VQA use

7

passes per model-example. VSR uses

8

passes because relation-swap probing is available. Routing adds two additional branches per model-output instance.

Field	Value
GPU type	NVIDIA RTX A6000 (48 GB)
Number of GPUs	3
CPU and memory	AMD EPYC 7543 32-Core Processor, 324GB
Inference framework	HuggingFace Transformers ( $\geq$ 4.40.0)
Batch size	1 (sequential per-sample inference)
Elapsed calendar time	$\approx$ 15 days

Table 42: Hardware and runtime used for the submitted experiments. Runtime depends on serving backend, batching, model implementation, and image preprocessing, so forward-pass counts in Table 41 remain the primary reproducibility quantity.

Artifact	Type	Use In Paper	License / Access Notes
HallusionBench	Benchmark dataset	Visual illusion, attribute error, grounded hallucination, source-profile diagnosis, and routing analysis	The official HallusionBench repository lists the artifact under the BSD-3-Clause license . We use the image-paired evaluation subset and do not redistribute raw images.
POPE	Benchmark dataset and evaluation code	Object-presence hallucination, blank-image pressure, abstention boundary check, and object-hallucination diagnosis	The official POPE repository provides the benchmark code and data. Its license file lists the artifact under the MIT License. We use released prompts and annotations for evaluation.
VSR	Benchmark dataset	Spatial relation verification and relation-swap alignment probing	The official Visual Spatial Reasoning repository lists the project under the Apache-2.0 license . We use zero-shot development examples with explicit spatial relations.
VizWiz-VQA	Benchmark dataset	Visual insufficiency, answerability diagnosis, and visual-fragility analysis	The VizWiz-VQA dataset page states that the work is licensed under a Creative Commons Attribution 4.0 International License . Since the images can contain user-provided visual contexts, we do not redistribute raw images.

Table 43: Dataset artifacts used in HalluPrism . We use public benchmark datasets and follow the corresponding license and access terms.

Artifact	Type	Use In Paper	License / Access Notes
Gemma 4	Open-weight MLLM family	One evaluated model family for Perturbation-Response measurement and routing analysis	Google AI documentation describes Gemma 4 as open-weight with responsible commercial use. Use remains subject to the applicable Gemma terms , Gemma Prohibited Use Policy , and the Gemma 4 model documentation .
Qwen3-VL	Open-weight MLLM	One evaluated model family for Perturbation-Response measurement, abstention scalarization, and routing analysis	The Qwen3-VL-8B-Instruct model card is used as the checkpoint reference. The model is listed with an Apache-2.0 license ; the Azure/Hugging Face catalog also lists Qwen3-VL-8B-Instruct as Apache-2.0.
LLaVA-Mistral	Open-weight MLLM	One evaluated model family for Perturbation-Response measurement and diagnostic controls	The llava-v1.6-mistral-7b-hf model card is used as the checkpoint reference. The model card metadata lists Apache-2.0 as the license.
LLaVA-Vicuna	Open-weight MLLM	One evaluated model family for Perturbation-Response measurement and diagnostic controls	The llava-v1.6-vicuna-7b model card is used as the checkpoint reference. The model card states that Llama 2 is licensed under the Llama 2 Community License .

Table 44: Model artifacts used in HalluPrism . We use the listed models for deterministic inference only and do not redistribute model weights.

Derived Artifact	Content	Release Boundary
Prediction files	Clean model answer, normalized answer, scalar confidence, correctness flag, and failure-family label	Can be released with dataset identifiers, subject to dataset terms.
Perturbation-Response files	$V$ , $L$ , $A$ , raw dominant source, normalized dominant source, and probe metadata	Can be released as derived metadata.
Perturbation metadata	Names and parameters for blur, crop, brightness, noise, blank-image replacement, grounding, and relation-swap probes	Can be released as code and metadata.
Routing files	Generic intervention output, matched intervention output, fix indicator, break indicator, and post-intervention accuracy	Can be released as derived metadata if original dataset terms permit.
Prompt catalog	Base prompts, diagnostic probes, entanglement probes, generic intervention, and matched intervention templates	Can be released in full.

Table 45: Derived artifacts produced by the HalluPrism pipeline. These artifacts support reproducibility without requiring redistribution of raw benchmark images or model weights.

Use Case	Appropriate Use	Boundary Or Risk
Model auditing	Compare scalar confidence with Perturbation-Response Signatures and failure-family labels.	Does not recover hidden causal mechanisms.
Selective answering	Use scalar scores only under explicit validation-to-deployment assumptions.	A threshold is a deployment guardrail, not an instance-grounded guarantee.
Correction routing	Use $(V,L,A)$ to choose visual enhancement, anti-prior prompting, or relation checking.	Routing is conditional and may still break correct answers.
High-stakes deployment	Use only with domain-specific validation, monitoring, and human oversight.	Not sufficient for medical, legal, financial, or accessibility-critical certification.
Human audit	Check whether normalized dominant sources are semantically plausible.	Human labels do not prove causal source separability.

Table 46: Responsible-use boundaries for HalluPrism . The method is intended for diagnosis and matched correction analysis, not for standalone safety certification.

Hallucination Detection in VLMs

Hallucination detection in vision-language models identifies claims that are unsupported by images, video, or other visual evidence.

See all 72 papers

HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models
Jul 23, 2026Weilin Jin, Mingyu Wang, Wenbo Li +5Hallucination Detection in VLMs Multimodal Hallucination
Peking University Beijing, China · Joy Future Academy Beijing, China
Uncertainty-Aware Decision Making in Multimodal Large Language Models
Aug 17, 2026Abderrahmene Boudiaf, Irfan Hussain, Sajid JavedMultimodal Large Language Models LLM Uncertainty Estimation
Khalifa University of Science and Technology, Abu Dhabi, United Arab Emirates
Mitigating Visual Hallucinations in Multimodal Systems through Retrieval-Augmented Reliability-Aware Inference
Jun 14, 2026Pratheswaran Hariharan, Haiping Xu, Donghui YanProbability Calibration Selective Prediction
Computer and Information Science University of Massachusetts, Dartmouth, MA · Mathematics and Data Science University of Massachusetts, Dartmouth, MA

Jul 23, 2026J/K move · Enter open · S save

HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models

Weilin Jin, Mingyu Wang, Wenbo Li, Haoyang Huang+4

Peking University Beijing, China · Joy Future Academy Beijing, China

Hallucination Detection in VLMs Multimodal Hallucination LLM Hallucination Detection Multimodal Model Interpretability VLM Hallucination Mitigation

Open paper AskPDF arXiv

Although Multimodal Large Language Models have achieved strong performance across a wide range of vision-language tasks, they still suffer from hallucinations, where model outputs become inconsistent with the visual content, textual context, or commonsense knowledge. Existing studies primarily address this problem through coarse-grained detection. However, these approaches often provide insufficient diagnostic information for understanding hallucination types and supporting downstream hallucination mitigation. To bridge this gap, we propose fine-grained hallucination diagnosis for MLLMs, a new unified task that jointly performs hallucination detection, classification, and interpretable explanation generation. We develop an automated data generation pipeline and construct HalluScope-30K, a large-scale diagnostic dataset covering eight sources and five task categories. Based on this dataset, we design a multi-granular joint reward function and train two diagnosis models, HalluScope-4B and HalluScope-8B, which achieve state-of-the-art performance on both the MHALO benchmark and our fine-grained hallucination classification benchmark. Notably, detection and classification are mutually beneficial under joint optimization. Furthermore, diagnosis-driven feedback experiments show that the fine-grained diagnostic explanations produced by our model effectively guide target models to correct their hallucinations, with full diagnosis substantially outperforming all baselines on both Qwen3-VL-8B-Instruct and LLaVA-1.5-7B. Our code, data, and models are available at https://github.com/wkinglin/HalluScope.

Abstract

Hallucination Detection in VLMs

Explore similar work

HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models

Uncertainty-Aware Decision Making in Multimodal Large Language Models

Mitigating Visual Hallucinations in Multimodal Systems through Retrieval-Augmented Reliability-Aware Inference