HallusionBench

Control-group diagnosis of language hallucination and visual illusion in LVLMs.

POPE

Polling-based object probing for detecting object hallucination in LVLMs.

FaithScore

Atomic image-fact verification for fine-grained hallucination detection.

HaELM

A reproducible local language-model evaluator for LVLM hallucinations.

AMBER

An LLM-free pipeline for detecting existence, attribute, and relation hallucinations.

GLSim: Detecting Object Hallucinations in LVLMs via Global-Local Similarity

A training-free detector that combines global scene similarity with local visual grounding for object hallucination detection.

Instruction Lens Score: Your Instruction Contributes a Powerful Object Hallucination Detector for Multimodal Large Language Models

A training-free object hallucination detector that uses calibrated local and instruction-context consistency scores.

Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lens

An attention-lens analysis identifies middle-layer signals for object hallucination detection and visual-attention adjustment.

ZINA: Multimodal Fine-grained Hallucination Detection and Editing

A span-level detector that classifies six hallucination error types and proposes grounded edits for MLLM outputs.

Seeing is Believing: Rich-Context Hallucination Detection for MLLMs via Backward Visual Grounding

VBackChecker detects hallucinations by grounding response claims backward to image pixels with rich contextual evidence.

Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback

A detect-then-rewrite pipeline uses fine-grained object, attribute, and relationship feedback to train HSA-DPO.

HalLoc: Token-level Localization of Hallucinations for Vision Language Models

HalLoc introduces token-level hallucination localization data and a concurrent low-overhead detector.

MHALO: Evaluating MLLMs as Fine-grained Hallucination Detectors

A fine-grained hallucination detection benchmark that evaluates MLLMs on recognition and token-level localization.

Detecting and Preventing Hallucinations in Large Vision Language Models

M-HalDetect provides fine-grained multimodal annotations and reward-model signals for hallucination detection.

Hal-Eval: a Universal and Fine-grained Hallucination Evaluation Framework for Large Vision Language Models

A unified discriminative and generative evaluator covering object, attribute, relation, and event hallucinations.

TLDR: Token-Level Detective Reward Model for Large Vision Language Models

TLDR assigns token-level rewards to expose hallucinated spans and support visual-language self-correction.

Detecting and Evaluating Medical Hallucinations in Large Vision Language Models

Med-HallMark and MediHallDetector provide hierarchical medical hallucination evaluation and fine-grained detection.

Unified Hallucination Detection for Multimodal Large Language Models

UNIHD unifies claim-level hallucination detection with auxiliary tools and introduces the MHaluBench meta-evaluation benchmark.

VL-Uncertainty: Detecting Hallucination in Large Vision-Language Model via Uncertainty Estimation

VL-Uncertainty uses semantic-equivalent perturbations and response uncertainty to detect hallucinations without extra labels.

Logical Closed Loop: Uncovering Object Hallucinations in Large Vision-Language Models

LogicCheckGPT probes object-to-attribute and attribute-to-object consistency in a training-free closed loop.

Structural Graph Probing of Vision–Language Models

Graph-based probes model neuron-correlation topology to expose structural signals associated with multimodal behavior and hallucination.

Lyapunov Probes for Hallucination Detection in Large Foundation Models

Lyapunov Probes use stability-constrained perturbation analysis to distinguish stable factual regions from hallucination-prone boundaries.

HALP: Detecting Hallucinations in Vision-Language Models without Generating a Single Token

HALP predicts hallucination risk before decoding by probing visual, vision-token, and query-token representations.

VADE: Visual Attention Guided Hallucination Detection and Elimination

VADE learns sequential patterns from raw visual attention maps for fine-grained hallucination detection and mitigation.

HALLUSHIFT++: Bridging Language and Vision through Internal Representation Shifts for Hierarchical Hallucinations in MLLMs

HALLUSHIFT++ extends internal distribution-shift analysis to hierarchical category, attribute, and relation hallucinations in MLLMs.

PAS: Prelim Attention Score for Detecting Object Hallucinations in Large Vision-Language Models

PAS measures attention on preliminary output tokens as a training-free, reference-free object hallucination signal.

MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs

MTRE aggregates early multi-token logits with likelihood-ratio evidence to estimate VLM response reliability.

Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT

ACT-ViT treats full layer-by-token activation tensors as image-like inputs for cross-LLM hallucination detection.

TruthPrInt: Mitigating Large Vision-Language Models Object Hallucination Via Latent Truthful-Guided Pre-Intervention

A latent-state detector identifies hallucinated object tokens and guides decoding toward a shared truthful direction.

Beyond the Global Scores: Fine-Grained Token Grounding as a Robust Detector of LVLM Hallucinations

A patch-level detector combines attention dispersion and cross-modal grounding consistency to localize hallucinated tokens.